You are on page 1of 35

University of California at Berkeley

School of Law
Rational Ignorance at the
Patent Office
by Mark Lemley
UC Berkeley School of Law
Public Law and Legal Theory
Working Paper No. 46
February 2001

UC Berkeley Public Law and


Legal Theory Working Paper Series
UC Berkeley School of Law, Boalt Hall, Berkeley, CA 94720-7200

Northwestern University Law Review, Vol. 95, No. 4

This paper can be downloaded without charge from the


Social Science Research Network Paper Collection at
http://papers.ssrn.com/paper.taf?abstract_id=261400

Copyright 2001 by Northwestern University School of Law


Northwestern University Law Review

Printed in U.S.A.
Vol. 95, No. 4

Essay
*

RATIONAL IGNORANCE AT THE PATENT OFFICE

Mark A. Lemley**
How much time and money should the Patent and Trademark Office
spend deciding whether to issue a patent? To judge by recent criticism
of the office from academics, industry leaders, and the press, the answer
is a lot more than it does now. The PTO has come under attack of late
for failing to do a serious job of examining patents, thus allowing bad
patents to slip through the system.1 The criticism is particularly strong in
specific industries, notably software and Internet business method patents, in which the PTO has arguably failed to respond quickly enough to
changing legal circumstances.2 Several solutions have been proposed,
*

Yes, I know its the Patent and Trademark Office, not the Patent Office. But nothing in this
Essay applies to the PTOs trademark operations, and besides being awkward, using the full title
might be misleading.
**
Professor of Law, Boalt Hall School of Law, University of California at Berkeley; of counsel,
Fish & Richardson P.C.
The comments I received on early drafts of this paper were even more helpful than usual, and I
have revised the paper quite substantially based on those comments. Thanks to John Allison, John
Barton, Bob Cooter, Alison Derbenwick, Rochelle Dreyfuss, Aaron Edlin, Jesse Fried, Wendy
Gordon, Rose Hagan, Justin Hughes, Mark Janis, Brian Kahin, Michael Katz, Jay Kesan, Dan Klerman, David McGowan, Peter Menell, Craig Nard, Cecil Quillen, Arti Rai, Pam Samuelson, Howard
Shelanski, Deepak Somaya, Yossi Spiegel, Jay Thomas, Hans Troesch, Keith Witek and participants
at the American Law and Economics Association meeting and the Olin Law and Economics Workshop at Boalt Hall for comments on an earlier draft, and to Frank Goldberg for outstanding research
assistance.
1
Complaints are legion. See, e.g., Julie E. Cohen, Reverse Engineering and the Rise of Electronic Vigilantism: Intellectual Property Implications of Lock-Out Programs, 68 S. CAL. L. REV.
1091, 1177-80 (1995); Andy Johnson-Laird, Looking Forward, Legislating Backward?, 4 J. SMALL
& EMERGING BUS. L. 95, 120-24 (2000); Jay P. Kesan & Marc Banik, Patents as Incomplete Contracts: Aligning Incentives for R&D Investment with Incentives to Disclose Prior Art, 2 WASH. U.
J.L. & POLY 23 (2000); Robert P. Merges, As Many as Six Impossible Patents Before Breakfast:
Property Rights for Business Concepts and Patent System Reform, 14 BERKELEY TECH. L.J. 577
(1999); John R. Thomas, Collusion and Collective Action in the Patent System: A Proposal for Patent Bounties (2000) (working paper, on file with author); Simson Garfinkel, Patently Absurd,
WIRED, July 1994, at 104; James Gleick, Patently Absurd, N.Y. TIMES MAGAZINE, Mar. 12, 2000, at
44; Lawrence Lessig, The Problem with Patents, STANDARD, Apr. 23, 1999, available at
http://www.thestandard.com/article/display/ 0,1151,4296,00.html; Greg Ahronian, Patenting the
Internet, Electronic Commerce, Bioinformatics, at http://www.bustpatents.com/index.html; Jeff Bezos,
An
Open
Letter
on
the
Subject
of
Patents,
http://www.amazon.com/exec.obidos/subst/misc/patents.html.
2
This criticism has a great deal of force with respect to software patents, where the PTO clearly
missed the ball for over a decade, failing to hire examiners skilled in the software arts or to allow
software engineers to practice before it, and failing to classify software prior art well. See, e.g.,
MARK A. LEMLEY ET AL., SOFTWARE AND INTERNET LAW 333-34 (2000); Cohen, supra note 1, at
1177-80; Julie E. Cohen & Mark A. Lemley, Patent Scope and Innovation in the Software Industry,
89 CAL. L. REV. 1, 42-45 (2001) (arguing that significant defects in the PTOs ability to examine
software patent applications remain unaddressed). By contrast, the PTO has reacted much more
quickly to the change in legal rules in 1998 that permitted the patenting of business methods. It developed new guidelines and started training examiners in 1999, and by 2000 had instituted special

NORTHWESTERN UNIVERSITY LAW REVIEW

but the common thread among them seems intuitively obvious: the PTO
should do a more careful job of reviewing patent applications and should
weed out more bad patents.
Much of the criticism of the PTO is well-founded, and I have no
intention of defending all of its policies3 or suggesting that all or even
most of the patents it issues are worthwhile.4 But the situation is more
complicated than this criticism suggests. Conducting a more thorough
examination of patent applications requires society to spend more time
and more money. Whether these increased resources would be spent
wisely depends on the return we get for that money. It is that return on
investment that I explore in a general way in this Essay.5
The essential insight of this Essay stems from the littleacknowledged fact that the overwhelming majority of patents are never
litigated or even licensed. Because so few patents are ever asserted
against a competitor, it is much cheaper for society to make detailed validity determinations in those few cases than to invest additional resources examining patents that will never be heard from again. In short,
the PTO doesnt do a very detailed job of examining patents, but we
probably dont want it to. It is rationally ignorant of the objective va-

review procedures for business method patents. See, e.g., John Schwartz, Online Patents to Face
Tighter Review, WASH. POST, Mar. 30, 2000, at E1.
3
For example, there are strong structural and psychological pressures on examiners to issue patents rather than reject applications, no matter how weak the alleged invention seems. Examiners
have astonishingly little time to spend on each applicationon average, a total of eighteen hours,
including the time spent reading the application, reading the submitted prior art, searching for and
reading prior art in databases accessible to the PTO, comparing that prior art to the application, writing an office action, reading and responding to the response to office action, iterating the last two
steps at least one and often more times, conducting an interview with the applicant, and ensuring that
the diagrams and claims are in form for allowance. Because so many applications arrive at the PTO
each year, examiners are rewarded for getting applications out the door. See Merges, supra note 1,
at 609. And because of a bizarre set of prosecution rules, it is impossible to reject a patent once and
for all. See 37 C.F.R. 1.60(b)(4), 1.62 (2000) (providing that after final rejection, applicant can
abandon and refile the same applicationcalled a continuationan unlimited number of times).
Further, examiners must write up reasons for rejection, but not reasons for allowance, giving them
more incentives to allow rather than reject an application. See, e.g., Thomas, supra note 1, at 21.
Finally, the examiner has the burden of proof in rejecting a patent application. See, e.g., In re
Oetiker, 977 F.2d 1443 (Fed. Cir. 1992); Thomas, supra note 1, at 21. Thus, the only way for an examiner to guarantee that an application is finally disposed of is to issue a patent. Examiners who
want credit for dispositions therefore have a strong incentive to issue patents to persistent applicants,
rather than to continue rejecting the applications. See Merges, supra note 1, at 590 (describing this
process).
There are other, less tangible incentives for examiners to issue patents in doubtful cases. The
PTO during the 1990s reengineered itself, declaring its mission to be helping our customers get
patents. This is an indefensible position for a quasi-judicial administrative agency that is trusted
with representing the public interest in deciding whether to issue patents. While the job of the PTO
is certainly to issue good patents, it is also to reject bad ones. The idea that applicants, rather than
the public at large, are the intended beneficiaries of the patent system cannot help but contribute to
the push to issue patents regardless of quality.
4
Recent evidence suggests that a large percentageperhaps as many as halfof the patents
that are eventually litigated to judgment are held invalid. See John R. Allison & Mark A. Lemley,
Empirical Evidence on the Validity of Litigated Patents, 26 AIPLA Q.J. 185 (1998).
5
While I have collected hard numbers for as many of the variables as possible, I wish to make
clear that my goal is to work out some back-of-the-envelope numbers that can be used to set general
policy.

95:#### (2001)

Rational Ignorance at the Patent Office

lidity of patents, in economics lingo,6 because it is too costly for the PTO
to discover those facts.
In Part I, I set out the salient facts about patent prosecution and its
costs. In Part II, I do the same with patent litigation and licensing, and
explore what happens to the vast majority of patents that are never litigated or licensed. In Part III, I explore what these numbers mean for
patent system reform. In Part IV, I respond to some likely objections to
the argument Ive presented. Finally, in Part V, I offer some policy proposals that stem from these ideas.
I. THE COST OF OBTAINING PATENTS
The number of patents issued in the U.S. has been increasing. It
gradually pushed past 100,000 per year during the 1980s and 1990s, and
skyrocketed to about 150,000 per year beginning in 1998. In 2001, patents in force are those issued after the same date in 1984, with the exception of a small number that had terminal disclaimers reducing their
terms,7 and a small number that had patent term extensions lengthening
their terms.8
TABLE 1: UTILITY PATENTS ISSUED PER YEAR9
Year of Issue
Number of Patents Issued
1999
153,493
1998
147,521
1997
111,983
1996
109,646
1995
101,419
1994
101,676

The data from the years before 1989 are substantially lower, but the
estimated patents for the immediate future are even higher. PTO critic
Greg Aharonian predicts the PTO will issue as many as 200,000 patents
in 2000.10 The total number of patents in force as of this writing is approaching two million.
Not all patent applications result in issued patents. Historically,
about sixty to sixty-five percent of patent applications have resulted in

The basic idea of rational ignorance is that any person will spend only a certain amount of time
or money to obtain a piece of information. If obtaining that information costs more than the information is worth, they will (or should) rationally choose to remain ignorant of it. See RALPH T.
BYRNS & GERALD W. STONE, ECONOMICS 433 (4th ed. 1989) (discussing rational ignorance); Mancur Olson, Rational Ignorance, Professional Research, and Politicians Dilemmas, in KNOWLEDGE,
POWER AND THE CONGRESS 130 (William H. Robinson & Clay H. Wellborn eds., 1991) (discussing
the idea in the context of citizens and their rational ignorance of public policies).
7
See 37 C.F.R. 1.321 (addressing rules regarding terminal disclaimers).
8
See 35 U.S.C. 155, 155A, 156.
9
U.S. Patent and Trademark Office, U.S. Patent Statistics, Calendar Years 1963-2000 (2001), at
http://www.uspto.gov/web/offices/ac/ido/oeip/taf/us_stat.pdf.
10
E-mail from Gregory Aharonian to Mark A. Lemley, Professor of Law, Boalt Hall School of
Law, University of California at Berkeley (Sept. 15, 2000) (on file with author); see also Barrett J.
Riordan, Whats Driving Patent and Trade Mark Application Filings?, 2000 EUR. INTELL. PROP.
REV. 349 ([T]he current rate of growth in the United States not only is high, but also appears to be
accelerating.).

NORTHWESTERN UNIVERSITY LAW REVIEW

issued patents, according to PTO data.11 Table 2 shows the number of


patent applications filed during the same period of years.
TABLE 2: UTILITY PATENT APPLICATIONS FILED PER YEAR12
Year of Application
Number of Applications Filed
1999
270,187
1998
243,062
1997
215,257
1996
195,187
1995
212,377
1994
189,857

Given the rate of increase, it is reasonable to project in future years a


minimum of 275,000 applications and 150,000 issued patents per year.
Prosecuting patents is expensive. There is some disagreement on
precisely how expensive it is, but the general range of costs for prosecuting a patent from start to finish (including application and various filing
fees paid to the PTO, and attorneys fees not only to prepare and file the
application, but to respond to office actions and continue prosecution
through to issuance or abandonment) appears to be $10,000 to $30,000
per patent.13 I have chosen a conservative average estimate of $20,000
per initial application taken through prosecution. Much of this cost is
front-loaded: it covers an attorneys time in meeting with the inventor,
writing the application, and writing patent claims, as well as a substantial
filing fee to the PTO. Other costs are incurred on a piecemeal basis as
prosecution progresses, and include both attorneys fees and PTO fees to
file each new piece of paper, up to and including the issuance of the patent itself. These cost averages include both patents that are ultimately is11

See Cecil D. Quillen, Jr. & Ogden H. Webster, Continuing Patent Applications and Performance of the U.S. Patent Office (2000) (working paper, on file with author). As Quillen and Webster
point out in their astonishing paper, however, these numbers are misleading because they include a
significant number of continuation applications and related refilings of previously-filed patent applications. I discuss their paper and its implications below.
12
U.S. Patent and Trademark Office, U.S. Patent Statistics, Calendar Years 1963-2000, supra
note 9. Note that because the average patent application takes over two and a half years to mature
into an issued patent, it is impossible to compare applications and issued patents year for year. See
John R. Allison & Mark A. Lemley, Whos Patenting What? An Empirical Exploration of Patent
Prosecution, 53 VAND. L. REV. 2099, 2118 (2000). One can get a sense for the likely number of
patents to issue in the next few years, however, by applying the 60-65% rule of thumb to the number
of applications filed in the last few years.
13
There is a great deal of variance in estimates of these numbers, and very little hard statistical
data. Part of the variance can be explained by the different degrees of complexity in different applications. Short applications with few claims that are allowed almost immediately obviously cost less
than long applications with multiple claims that spend a great deal of time in prosecution. Some estimates follow: Erwin F. Berrier, Jr., Global Patent Costs Must Be Reduced, 36 IDEA 473, 476-77
(1996) (describing cradle-to-grave costs of prosecution as $14,370 for a typical application); Jon
D. Grossman & Eric Oliver, A Step-by-Step Guide to Prosecuting Business Method Patents, 17:3
COMP. L., Mar. 2000, at 6, 9 ([T]he median cost of preparing and prosecuting a utility patent application approaches $15,000 in legal fees alone. (citing the 1997 American Intellectual Property Law
Association (AIPLA) Economic Survey of Patent Lawyers)); Wayne M. Kennard, Obtaining and
Litigating Software Patents, 430 PLI/PAT 193, 208 (1996) (estimating the average cost for preparing
patent applications in the software field to be $10,000 to $30,000, and the average cost of prosecuting them to be another $10,000 to $20,000).
Note that none of these estimates include either appeals or interferences, which obviously raise
the cost a great deal.

95:#### (2001)

Rational Ignorance at the Patent Office

sued and patent applications that are ultimately rejected by the PTO
without being revived.
A significant percentage of the total applications filed in the PTO,
however, are not initial applications but continuing prosecution applications.14 Quillen and Webster collect data from the PTO showing that in
the 1990s, 28.4% of all applications filed were continuing prosecution
applications.15 Continuing prosecution applications are much less expensive than initial applications, because in many cases the specification has
already been written and much of the prior art has already been researched. I will use a conservative cost estimate of $5,000 per continuation application.
With these numbers, we are in a position to calculate some approximate annual costs of patent prosecution.16 Of the 275,000 applications filed per year, 28.4%, or 78,100, are continuation applications. The
cost of 196,900 regular applications prosecuted through to issuance or rejection at $20,000 per patent totals $3.94 billion. An additional 78,100
continuing patent applications at $5,000 per patent costs $391 million.
This gives us a total annual cost of $4.33 billion for domestic patent
prosecution.
What does this money buy? The answer is, surprisingly little actual
assessment of whether a patent should issue. The patent prosecution
process is ex partethe only participants are the applicant seeking a patent and the examiner, who is both judge and devils advocate.17 While
patent applicants must submit to the PTO relevant prior art of which they
are aware,18 they are under no obligation to search for prior art, and most
do not. The examiner, then, has the burden of reading the application,
searching for and identifying the relevant prior art, reading the relevant
prior art, deciding whether the application should be allowed by comparing the claims to the prior art, and writing an Office Action explaining
the reasons why any claims are rejected. After the applicant writes a response to the examiners evaluation, this process will normally happen
again, and may happen several more times. The examiner may also conduct an interview with the applicant to discuss allowance in person or
over the phone. Finally, there are technical matters that the examiner
must identify and attend to before the patent application is in condition
for allowance. The total average time the examiner spends on all these
tasks over the two- to three-year prosecution of the patent is eighteen

14
By the use of this term here, I mean to include divisionals, continuations, and continuation-inpart applications.
15
See Quillen & Webster, supra note 11.
16
I should emphasize that these are costs in the United States alone. Patent applicants that want
protection outside the United States must pay radically more, as they must replicate the prosecution
process in many different countries.
17
The patent law was recently amended to provide for third-party participation in the reexamination of patents that the PTO has already issued. See 35 U.S.C. 311-318 (West Supp. 2000).
However, virtually no one is expected to use this system, because doing so precludes you from challenging the validity of a patent in later litigation. 315(c); see Mark D. Janis, Inter Partes Patent
Reexamination, 10 FORDHAM INTELL. PROP. MEDIA & ENT. L.J. 481, 493 (2000) (arguing that Congress has exacted a very high price for participation in an inter partes reexamination).
18
37 C.F.R. 1.56.

NORTHWESTERN UNIVERSITY LAW REVIEW

hours.19 It is not surprising, therefore, that the PTO issues many patents
that would have been rejected had the examiner possessed perfect
knowledge. This is particularly true since much of the most relevant
prior art isnt easy to findit consists of sales or uses by third parties
that dont show up in any searchable database and will not be found by
examiners in a hurry. Indeed, in litigated cases that actually result in a
final judgment on validity, issued patents are held invalid forty-six percent of the time.20
II. WHAT PATENTEES DO WITH THEIR PATENTS
To evaluate whether additional time and money investigating patent
applications would be well spent, we also need an idea of what patentees
actually do with their patents. Surprisingly, hard data on this question
are hard to come by. The traditional incentive story relied upon by intellectual property scholars assumes that people seek patents to obtain exclusive rights to a technology, and that they use those patents either to
exclude competitors from the market or to obtain licensing revenue in
exchange for permitting the use of the patented technology.21 This story
is not only incomplete, but dramatically so.22 The limited data we have
suggest that the overwhelming majority of patents are neither litigated
nor licensed.
A. Litigation
Of the roughly two million patents currently in force,23 only a tiny
number are the basis for lawsuits each year. About 1,600 patent lawsuits
are filed each year,24 involving at most perhaps 2,000 different patents.25
19

See, e.g., Brenda Sandburg, Speed Over Substance?, INTELL. PROP. MAG., Mar. 1999 (estimating eighteen hours on average; examiners may spend more time on complex technologies);
Thomas, supra note 1, at 10 (estimating sixteen to seventeen hours); Interview with Q. Todd Dickinson, Director of the U.S. PTO, in Berkeley, Cal. (Apr. 2000); cf. Patent Nonsense: The Knowledge
Monopolies, ECONOMIST, Apr. 8, 2000 ([P]atent examiners spend only eight hours on a patent, on
average.).
I should make it clear that this is an average across all industries, and that there may be substantial variation in the hours spent from one industry to another. See Interview with Q. Todd Dickinson, supra (stating that the hours spent range from eight per patent in some art units to thirty-two in
other art units).
20
Allison & Lemley, supra note 4, at 205-06.
21
See, e.g., Mark A. Lemley, The Economics of Improvement in Intellectual Property Law, 75
TEX. L. REV. 989 (1997) (recounting the traditional story and citing sources).
22
See Mark A. Lemley, Reconceiving Patents in the Age of Venture Capital, 4 J. SMALL &
EMERGING BUS. L. 137 (2000) (questioning the traditional incentive story and seeing a need for research into the real causes of innovation).
23
See supra notes 7-12 and accompanying text.
24
The exact data for the years 1995-1999 can be found in the Derwent Litalert database, available at http://www.derwent.com/intellectualproperty/litalert.html. The data that follow were compiled as of June 1, 2000, and involve cases labeled patent.
Year
Number of Cases Filed
1999
1,652
1998
1,730
1997
1,731
1996
1,514
1995
1,258
Strictly speaking, I should consider the number of cases likely to be filed 5-6 years into the future,

95:#### (2001)

Rational Ignorance at the Patent Office

The overwhelming majority of these lawsuits settle or are abandoned before trial. Only about one hundred cases per year (and 125 patents) actually make it to trial.26 Based on these numbers, it is reasonable to
estimate that at most only about two percent of all patents are ever litigated, and less than two-tenths of one percent of all issued patents actually go to court.27
When patent litigation does occur, it is expensive. The American
Intellectual Property Law Association reports, based on a survey of its
members, that the median cost of patent litigation to each side is
$799,000 through the end of discovery, and $1,503,000 through trial and
appeal.28 Estimating a total annual cost requires us to compute the cost
of cases that go to trial ($1,503,000 x 100 cases x 2 parties = $301 million) and the cost of cases that are filed but settle before trial.29 In the
latter case, multiplying $799,000 times the 1,500 cases annually filed but
not taken to trial times two parties gives us a cost estimate of $2.4 billion. In practice, however, some cases will settle early and therefore not
incur this cost30 (though based on my anecdotal experience, cases are
more likely to settle late in the litigation process, when the cost and uncertainty of trial is looming). To account for this, let us assume that the
cost of cases settled before trial is only three-fourths of $2.4 billion, or
$1.8 billion. This $1.8 billion plus $301 million gives us a total annual
amount spent on patent litigation of $2.1 billion.31 It is worth noting that
since patent lawsuits arent normally filed immediately after the patent issues. Cf. Allison & Lemley, supra note 4, at 236 Table 12. The trends suggest that more cases will be filed in the future.
25
Some patent lawsuits involve more than one patent. Two recent studies found about 1.25 patents were involved per lawsuit. See, e.g., Allison & Lemley, supra note 4, at 205 (presenting data
that show 299 patents in 239 cases actually resolved by opinion between 1989 and 1996, or 1.25 patents per cases); Kimberly Moore, Judges, Juries and Patent CasesEmpirical Evidence to Peek Inside the Black Box, 99 MICH. L. REV. 365 (2000) (providing data for all cases that went to trial
between 1983 and 1999 that show 1733 patents in 1355 cases, or 1.28 patents per case). If about
1,600 different cases are filed per year, these data suggest that those cases would involve approximately 2,000 patents, unless there is some reason to believe cases that actually get resolved are likely
to involve systematically more (or fewer) patents than those cases that settle. But some patents are
doubtless included in more than one case, so the actual number of different patents litigated from
each cohort should be somewhat less than 2,000.
26
See Moore, supra note 25, at 24 (stating that the number of trials per year from 1988 to 1998
ranged from a low of 86 to a high of 108).
27
If anything, these numbers are on the high side, because many patents that do go to court are
litigated in more than one case. Thus, the total number of different patents litigated may actually be
somewhat smaller.
Interestingly, the numbers appear to vary somewhat by industry. See Josh Lerner,
Patenting in the Shadow of Competitors, 38 J. L. & ECON. 463 (1995) (study showing that 6% of
biotechnology patents were involved in litigation).
28
AIPLA Report of Economic Survey (of U.S. IP Practitioners) (1999), cited in Craig P. Opperman, Computer Technology Patents (with an Emphasis on Internet & E-Commerce Related Patents), 590 PLI/PAT 1039, 1047 (2000). The average cost of patent litigation is likely higher, because
some extremely expensive cases will raise the average above the median. The study does report that
the seventy-fifth percentile of patent cases cost $1,503,000 to take through discovery, and
$2,510,000 through the end of the case.
29
This is only an estimate, because I have access only to the median cost and not the cost distribution. It is likely that these numbers therefore understate the costs of patent litigation somewhat,
because extremely expensive cases arent fully accounted for by the median number.
30
Recall that the $799,000 number represents a median only for cases that continue through the
end of the discovery period, not for all cases settled short of trial.
31
An earlier and smaller estimate comes from Josh Lerner, who studied cases filed in 1991 and

NORTHWESTERN UNIVERSITY LAW REVIEW

this number is only the cost of legal fees, and does not include the cost of
judgments paid, which would raise defendants costs but lower plaintiffs
correspondingly. It also does not include indirect social costs such as judicial resources or the value of the time lost by corporate employees involved in the case.
Not all of this cost, of course, is attributable to litigating the validity
or enforceability of the patent. Virtually every patent infringement lawsuit includes a claim that the patent is either invalid or unenforceable due
to inequitable conduct (or commonly both). But these cases also involve
infringement allegations that must be litigated, and commonly include
ancillary issues as well (e.g., damages and license questions). Validity
requires a great deal of attention, however. In contrast to the eighteen
hours an examiner will spend on a patent from start to finish, lawyers and
technical experts will spend hundreds and perhaps even thousands of
hours searching for and reading prior art, poring over the specification
and prosecution history, and preparing and defending invalidity arguments. A rough estimate may be that half of the cost of patent litigation$1.05 billion per yearis attributable to disputes over the validity
or enforceability due to inequitable conduct of the patents in suit.32 This
leaves the remaining cost for infringement, license, antitrust, damages,
willfulness and the related non-patent issues that are often litigated in
patent cases.
B. Unlitigated Patents
What happens to the rest of the patents? A surprisingly large number
of issued patents lapse for failure to pay required maintenance fees.33
Payment of these fees, which are relatively low,34 is a prerequisite to
bringing a patent lawsuit: failure to pay maintenance fees is effectively
an abandonment of the patent.35 The evidence presented in Table 3 suggests that nearly two-thirds of all issued patents lapse for failure to pay
maintenance fees before the end of their term: nearly half of all patents
are abandoned in this way before their term is half over. Most of these
patents arent litigated or licensed during the short time they are in
force.36
determined that they would cost about $1 billion. See Josh Lerner, Patenting in the Shadow of Competitors, 38 J.L. & ECON. 463, 470 (1995).
32
It is possible that the amount of money litigants spend in lawsuits relates more to the perceived importance of the case and the litigants financial condition than to the issues. Thus, if we
were to remove validity issues entirely from litigation, it may be that the parties would simply spend
more money litigating the other issues.
33
Maintenance fees are due in increasing amounts at periods three and a half years, seven and a
half years, and eleven and a half years after the patent issues. 35 U.S.C. 41(b) (West Supp. 2000).
For a pioneering study of maintenance fees and patent value, see Jean O. Lanjouw et al., How to
Count Patents and Value Intellectual Property, 46 J. INDUS. ECON. 405 (1998).
34
The fees are $830 at three and a half years, $1,900 at seven and a half years, and $2,910 at
eleven and a half years. 35 U.S.C. 41(b).
35
The issue is more complicated, because it is sometimes possible to reinstate patents that have
lapsed for failure to pay maintenance fees. See, e.g., Centigram Comm. Corp. v. Lehman, 862 F.
Supp. 113 (E.D. Va. 1994).
36
Some patents that lapse may nonetheless have been used, and may merely have become irrelevant with the passage of time. But patents that are litigated or licensed are generally not those
that expire for failure to pay the fairly modest maintenance fees.

95:#### (2001)

Rational Ignorance at the Patent Office

Obviously, though, many patents that do remain in force never get


litigated. Some of these patents are licensed for royalties without litigation. Surprisingly, it does not appear that anyone knows precisely how
many patents are licensed for royalties.37 There are reasons to believe,
however, that the number is not large.
TABLE 3: PERCENTAGE OF PATENTS FOR WHICH
MAINTENANCE FEES WERE PAID38
Age of Patent39
Percentage Paid
0-1 year
100%
1-2
100%
2-3
100%
3-4
100%
4-5
82.56%
5-6
81.68%
6-7
80.77%
7-8
78.78%
8-9
57.43%
9-10
56.53%
10-11
56.14%
11-12
56.23%
12-13
37.05%
13-14
36.78%
14-15
36.00%
15-16
35.73%
16-17
36.03%

First, many patentees engage in defensive patenting, obtaining


patents to stake their claim to an area of technology in hopes of preventing other companies from suing them. Indeed, there is anecdotal evidence that at least among high-technology and start-up companies, the
primary purpose of patents is defensive.40 Licensing patents for royalties
is correspondingly uncommon in many industries in which all the major
37

Jim Pooley has proposed that all license agreements be recorded in the PTO. James H.A.
Pooley, The Trouble with Patents, CAL. LAW. (2000); cf. Ronald H. Coase, in ESSAYS ON
ECONOMICS AND ECONOMISTS (1975) (proposing a national contracts database). In addition to being a good idea for the reasons he suggests, this proposal would have the collateral benefit of improving our knowledge about the patent system dramatically. In the absence of such a rule, this is an
area that is ripe for research.
38
See E-mail from Paul Harrison, USPTO, to Franklin Goldberg, research assistant to Mark
Lemley (June 6, 2000) (on file with author); see also Lanjouw et al., supra note 33, at 405. Data
made available from Bill Brown, USPTO Maintenance Fee Office. The data are from 1998.
39
Maintenance fees are due at 3 and a half years, 7 and a half years, and 11 and a half years after the patent issues. See 35 U.S.C. 41(b). Thus, for the first three years no fees are due, and the
percentage is 100%.
40
See, e.g., John H. Barton, Reforming the Patent System, 287 SCIENCE 1933 (2000) (arguing
that reducing the number of patents would help to solve the problem of defensive patent portfolios); Lemley, Reconceiving Patents, supra note 22, at 143 (One of the major reasons that companies get patents is that theyre afraid that their competitors have them, and they dont want to be the
only one left who doesnt have the ability to play in this game.); Scott Herhold, Patents Emerge as
Online Battleground, S.J. MERCURY-NEWS, July 17, 1999 (quoting a venture capitalist as saying
None of my companies seek patent protection because they actually think it will protect them from
competition. . . . Rather, they seek patents to protect themselves from other people who have patents.).

NORTHWESTERN UNIVERSITY LAW REVIEW

players have large patent portfolios. Patent licensing in such an industry


has a very different character from the typical model of licensing for royalties. Large companies tend to come to the table with hundreds of patents on each side, relying on volume rather than quality in a sort of
patent arms race.41 While some cross-licensing deals in such industries
are royalty-bearing, it is more common for companies to agree to royalty-free cross-licenses, in which each party gets the freedom to make
products but does not have to pay the other. Similarly, in many hightechnology industries patent rights are waived (or licensed on a royaltyfree basis, which amounts to the same thing) because the patented technology is adopted as an industry standard.42
Second, patentees who want to license their patents for royalties are
typically parties with asymmetric stakesthey are individuals who dont
sell products, licensing shops whose primary output is patents, or older
companies that are no longer major players in the marketplace. Parties in
these situations have no need to trade patents in the patent arms race
described above.43 These patentees often approach an entire industry at
once in an effort to maximize revenue. While some of those patents may
be successfully licensed without litigation, it is more likely that at least
one company they approach will opt to fight rather than take a license. If
even one company does so, the patent is included in the litigation percentages cited above.44 Further, litigation may be a useful settlement
strategy, forcing the other side to the bargaining table. Indeed, that appears to be what litigation normally is, since the overwhelming majority
of patent lawsuits settle, presumably with some form of licensing deal.
Third, many companies obtain patents for reasons totally unrelated
to litigation or licensing. Companies may patent broadly to hedge their
bets if they are uncertain what patents are likely to have value ex post.
Other companies, especially start-ups, obtain patents as a financing
tool.45 Venture capitalists use client patents (or more likely, patent applications) as evidence that the company is well managed, is at a certain
41

See, e.g., Barton, supra note 40, at 1933 (suggesting that this practice is common in other industries as well); Bronwyn Hall & Rosemarie Ham Ziedonis, The Patent Paradox Revisited: Determinants of Patenting in the U.S. Semiconductor Industry, NBER Working Paper 7062 (1999)
(discussing such cross-license patterns in the semiconductor industry).
At a conference in November 2000, the former general counsel of a major semiconductor company said that at his company, intellectual property lawyers would index their patents against competitors products, but wouldnt assert those patents in negotiations unless and until the competitor
approached them first. My sense is that this anecdotal evidence is quite representative of the approach in many industries.
42
See Mark A. Lemley, Antitrust, Intellectual Property Rights and Standard-Setting Organizations (2001) (working paper, on file with author) (discussing organization rules that require licensing
of intellectual property on favorable or even royalty-free terms).
43
Thus, Jerome Lemelson is famous for having licensed his patents aggressively, and Texas Instruments is the most aggressive licensor of patent in the semiconductor industry. Lemelson did not
make any products himself, and therefore didnt need cross-licenses from anyone. Texas Instruments, while still a player in many markets, litigated primarily in the area of large-scale integrated
circuits, in which it did not have significant sales by the time of the lawsuits.
44
Thus, both Lemelson and Texas Instruments licensed their patents without litigation on many
occasions, but both have also been involved in multiple lawsuits over those patents.
45
See generally Charlotte H. Copperthite & Michael J. Lerner, Creative Use of IP Portfolios
Helps Secure Financing, NATL L.J., May 24, 1999 (discussing the role of IP rights in biotechnology
financing).

10

95:#### (2001)

Rational Ignorance at the Patent Office

stage in development, and has defined and carved out a market niche.46
Established companies may patent out of inertia, to maintain a reputation
as a market leader, or simply for the marquee value of selling a product
with patented technology.47 Individuals employed by those companies
may pursue a patent for its resume value, for the cash rewards that many
companies pay, or simply because it is a routine and established policy.48
Finally, some patents are issued, especially to individuals, not because
they have any market value but simply because the applicant really wants
a patent on his invention, no matter how little commercial value it is
likely to have.49
Even if a patent is obtained with an eye towards litigation or licensing, it may not be used in that way. Some patents get lost in a corporate
bureaucracyby the time the patent issues, it may be largely forgotten
by the licensing or litigation departments of a large corporation, or ignored by the general counsel of a smaller company with other things on
her mind.50 Other patents, even if filed with an intent to license or sue,
may simply turn out not to be useful, either because the claims are too
narrow or because by the time the patent issues the industry has moved
in an unanticipated direction. There is certainly a strong sense in the patent community that many issued patents, even potentially valuable ones,
sit around collecting dust.51
In terms of numbers, there is further work to be done. It is surprising that we dont have a very good idea of how many patents are actually
licensed for revenue. My sense, however, is that a relatively small percentage of the 150,000 or so patents issued each year are actually li46

Lemley, Reconceiving Patents, supra note 22, at 143-44; cf. Samuel Kortum & Josh Lerner,
Does Venture Capital Spur Innovation? (1998) (NBER working paper W6846, on file with author)
(finding that venture capital funding positively correlated with patenting).
47
See Ann Bartow, Separating Marketing Innovation from Actual Invention: A Proposal for a
New, Improved, Lighter, and Better-Tasting Form of Patent Protection, 4 J. SMALL & EMERGING
BUS. L. 1 (2000) (describing the widespread use of patents for such purposes and identifying examples).
48
At a conference in November 2000, the chief patent counsel for a biotechnology company
suggested that the demand in many technology-driven corporations for metrics to measure employees and departments drove individuals and groups within those companies to set numerical quotas for patents they would acquire.
49
Indeed, there is an entire book devoted to oddball patents on things like ear-flattening devices and thumb-wrestling rings that are unlikely ever to be licensed for royalties. See RICK
FEINBERG, PECULIAR PATENTS (1994).
50
Sam Oddi reports old data that many patents did not cover products in use at all:
In an industry survey conducted in the early 1950s, it was concluded that: The overall utilization rate of patents in current use, used in the past, and reported about to be used is 57 to 58%.
The survey consisted of over 1,000 questionnaires returned by companies concerning their use
of patents. The study further concluded that the patent utilization rate is significantly higher
for smaller companies (over 75%) compared to larger companies (close to 50%).
A. Samuel Oddi, An Uneasier Case for Copyright Than for Patent Protection of Computer Programs, 72 NEB. L. REV. 351, 426 n.310 (1993) (citations omitted) (quoting Barkev S. Sanders et al.,
Patent Acquisition by Corporations, 3 PAT. TRADEMARK & COPYRIGHT J. RES. & EDUC. 217, 239
(1959)). See also British Technology Group, IPR Market Benchmark Study, http://www.btgplc.com
(1998) (a survey of companies shows that only 15% had no unused patents in their portfolio, while
24% said they had over 100 unused patents, and 12% had more than 1,000 unused patents).
51
Thus, Kline and Rivette call their book on patents Rembrandts in the Attic because even those
who are holding onto a valuable asset often dont know it. KEVIN G. RIVETTE & DAVID KLINE,
REMBRANDTS IN THE ATTIC (2000).

11

NORTHWESTERN UNIVERSITY LAW REVIEW

censed to third parties in exchange for royalties. As we have seen, only


about 1.5% of all patents are litigated at all. The total number of patents
licensed for royalties without even a complaint being filed is likely
somewhat higher, but I suspect the total number of patents litigated or licensed for a royalty (as opposed to a cross-license) is on the order of five
percent of issued patents.52
The cost of licensing without going to court is also dramatically
lower than the cost of litigation. Hard data are again difficult to find, but
a reasonable estimate of the cost of negotiating a license might be
$50,000 per licensee per patent.53 Again, not all of this amount will be
attributable to validity issuesthe parties will also debate whether the
patent covers the accused product, and what the appropriate royalty rate
will be. Let us assume the validity investigation (normally requiring the
accused infringer to obtain a written opinion of counsel) and debate in an
average royalty-bearing patent license costs half of the total, or $25,000.
We then need to know how many licensees pay this cost for any given
patent. Some patents are licensed only to one party; others may be licensed to a host of companies. But the more companies a patentee approaches, the more likely the patent will end up in at least one lawsuit.
Further, patents asserted against a whole industry are likely to provoke
concerted responses. Accused infringers may turn to a law firm or
search firm that has already written opinions on this patent, reducing the
cost of a validity evaluation. To account for these facts, I have assumed
that the actual licensing cost to the industry is twice the cost of an individual license: $100,000 total, with $50,000 spent on validity. This gives
us a total annual licensing cost outside of litigation for 5250 patents,54 at
$100,000 per patent, of $524 million. The amount attributable to validity
for 5250 patents, at $50,000 per patent, is $262 million.
The licensing numbers are much more of an estimate than the numbers for either prosecution or litigation. Because I dont have hard numbers, in the calculations that follow I will work with a range of possible
numbers. In any event, the important lesson is that the overwhelming
majority of patents are never used in a way that calls their validity into
question.
52
This is obviously an important number for this article, and I am troubled that it is only an estimate. In Subpart IV.A, I relax this assumption to see how it affects the results.
For what its worth, while several academics thought the 5% number was too low, some experienced patent prosecutors told me it was too high. One former patent prosecutor for a major technology company said that his companys rule of thumb was that only 1% of their patents were worth
licensing. For other estimates roughly in line with my own, see Steven C. Carlson, Patent Pools and
the Antitrust Dilemma, 16 YALE J. REG. 359, 385 (1999) (Most infringed patents are not worth defending in court . . . Even in the pharmaceutical industry, where patents are most valuable, eight out
of ten patents typically produce no value for their holders.); Hearings on Global and InnovationBased Competition Before the Fed. Trade Comm'n 55 (1995) (statement of F.M. Scherer, Professor,
Harvard Univ.).
53
Some empirical work in the international technology-licensing context, where the costs are
likely to be higher, sets a value of about $100,000 per transaction. See FAROK J. CONTRACTOR,
INTERNATIONAL TECHNOLOGY LICENSING: COMPENSATION, COSTS, AND NEGOTIATION 105 (1981).
This number is likely to be significantly higher than the cost per patent of domestic licensing, both
because international deals tend to be more expensive and because many transactions involve more
than one patent.
54
150,000 patents per year times 3.5% (5% licensing rate, less the 1.5% that involve litigation).

12

95:#### (2001)

Rational Ignorance at the Patent Office

III. HOW TO IMPROVE THE PATENT SYSTEM


Suppose we agree with the host of commentators suggesting that
there are too many bad patents in the system, and we want to take steps
to weed out some of those patents. How should we go about doing it?
One way to do itthe method suggested by some advocates of patent reform55would be to have PTO examiners spend more time examining patents and, in particular, more time searching the prior art. More
time means more money will be spent by both the PTO (to hire additional examiners) and by patent applicants (for additional legal fees to respond to the new rejections these additional examiners will doubtless
issue). Doubling the amount of time examiners spend reviewing a patent
from eighteen hours to thirty-six hours might seem a reasonable place to
start. Doing so wouldnt double the cost of prosecution, because much
of that cost is incurred drafting the initial application, and those costs
would remain unchanged. But it might mean a fifty percent increase in
the cost of prosecution, from an average of $20,000 per original applications to $30,000 (and from $5,000 to $7,500 for continuing prosecution
applications).56 It should also mean that fewer patents issue, though the
size of the drop is hard to estimate. The reduction will depend on the
quality of the patents that currently issue and how many of the bad patents that currently issue can be smoked out merely by adding a few more
hours to an examiners evaluation. Let us assume the number of total issued patents would drop by ten percent to 135,000 per year as a result of
truly final rejections of more applications by the PTO.57 Finally, a more
comprehensive examination process might deter some people from filing
55

This argument is particularly common in the popular press. See, e.g., Gleick, supra note 1, at
44; Marc J. Lane, Talking Entrepreneurship: Patent Offices Monopoly Game Hurting Innovation,
CRAINS CHI. BUS., June 5, 2000; Philip E. Ross, Patently Absurd: Technology and Gamesmanship
Have Overwhelmed the U.S. Patent Office. How to Fix It?, FORBES, May 29, 2000. Members of
Congress have also made suggestions along these lines. See Victoria Slind-Flor, Bar Reacts to Bezos Patent Reform Plan, NATL L.J., Mar. 27, 2000 (quoting Representative Coble: If everyone
would keep their grubby hands off the PTOs fees, the agency could hire and retain even more examiners to ensure that only quality patents are issued.). Patent examiners themselves have also taken
this position. See Patent Reform and the Patent and Trademark Office Reauthorization for Fiscal
Year 2000 Before the Subcomm. on Courts and Intellectual Prop. of the House Comm. on the Judiciary, 106th Cong. 88 (1999) (statement of Ronald J. Stern, President, Patent Office Profl Assn)
(Patent examination needs to be enhanced by providing more time for examination.). Indeed,
Bruce Lehman, then Commissioner of the PTO, himself took that position back in 1994, explaining
[w]e are going to have to re-engineer the Patent and Trademark Office so that we give [examiners]
more time and more resources. That has a price tag. George Leopold, Congress, PTO Ponder
Ways to Streamline Operation, ELECTRICAL ENG. TIMES, Aug. 1, 1994.
56
Some of this increase will take the form of increased PTO fees needed to hire more examiners. PTO fees might be expected to roughly double to account for spending twice as much time per
patent. Another component of the cost will be attorneys fees needed to respond to the additional
office actions examiners will generate, the necessity of drafting and explaining more claim amendments, and the additional work that will be generated for lawyers by a more thorough PTO prior art
search that will presumably generate more prior art to evaluate and argue over.
57
I actually suspect that this estimate, which would mean a proportionately greater drop of over
20% in the number of bad patents, is quite high. Quillen and Webster demonstrate that the PTO in
its eighteen-hour review actually issues 97% of all the applications it receives. See Quillen & Webster, supra note 11, at 9, tbl.6. To think that adding eighteen more hours will more than quadruple
the effective rejection rate seems implausible. But I will use this number to be conservative.

13

NORTHWESTERN UNIVERSITY LAW REVIEW

applications at all, both because of the increased costs and because of the
decreased likelihood of actually getting a patent. To account for this, I
have assumed that the number of applications would reverse its significant upward trend, and in fact would fall by about 10% to 247,500 under
such a system. Thus, the total costs of prosecution under the new system
would be 177,200 issued patents at $30,000 per patent ($5.32 billion),
plus 70,300 continuing prosecution applications58 at $7,500 per patent
($530 million), for a total cost of $5.85 billion. This is an increase of
$1.52 billion in the cost of prosecution.59
How much money would this save on the back end? As we discussed in Part II, the total cost of patent litigation is $2.1 billion, and the
total cost of licensing outside of litigation is $524 million.60 Thus, $2.62
billion is spent each year on licensing and litigation. As noted above, it
is reasonable to expect the enhanced examination system will prevent
about ten percent of those patents from issuing.61 Thus, this cost should
drop by ten percent, or $262 million, much less than the $1.52 billion we
are spending to achieve this outcome.
An alternative proposal would require patent applicants to conduct a
prior art search before filing their application and to disclose the results
of that search to the PTO. This approach would have similar effects.
Prior art searches cost somewhere between $5,000 and $7,000 each on
average, though the actual price will differ depending on the complexity
of the technology and how crowded the field is. A prior art search also
requires the patentee to disclose more art, and therefore argue about that
art with the examiner. Thus, it will increase legal fees as well. We can
assume that requiring prior art searches would add between $8,000 and
$10,000 to the cost of patent prosecution in most cases.62 If we assume
58

275,000 applications less 10% = 247,500 applications. 125,000 would issue as patents, and
the rest would go abandoned.
59
There will likely be indirect costs as well. A longer and more involved examination will
likely lead to more initial rejections of all patent applications, even those that ultimately issue as patents. Thus, we might reasonably expect total prosecution times to increase under this strengthened
examination system, even if many more examiners are hired to relieve the burden on those presently
employed. Longer prosecution times in turn impose costs not only on the patentee, whose right of
exclusivity is delayed, but also on others in the industry, who may be proceeding with their own development plans in ignorance of a pending patent application that covers their product.
60
I have used the full costs here, not the percentage attributable to validity issues, because patent applications that do not issue under the more detailed examination regime cant be litigated or
licensed at all, and so none of that cost is incurred.
61
It is worth asking whether the patent applications that will not issue under this system would
be more or less likely than average to be the patents that would have ended up in litigation. On the
one hand, patentees are unlikely to spend money to litigate their weakest patents, so one might expect that the poor-quality patents that could have been knocked out by a slightly better examination
process wouldnt be litigated at all, or at least wouldnt make it very far. On the other hand, it is
possible that the patents that are invalidated under this stricter examination system are those with the
broadest claims, and that those broad claims are more likely to be asserted in litigation. The text
makes neither assumption, and instead simply assumes that a 10% reduction in the number of issued
patents will translate into a 10% reduction in the number of litigated and licensed patents. Varying
this assumption will of course vary the numbers.
62
This is true only if the patentee doesnt conduct a prior art search already. But most dont,
and in particular experienced patent prosecutors are less likely to conduct a prior art search than novices. In part this is because the PTO requires that patent applicants disclose information of which
they are aware, but does not require a search, so any search that is done may make it harder to get a
patent. See FMC Corp. v. Hennessy Indus., 836 F.2d 521, 526 n.6 (Fed. Cir. 1987) (As a general

14

95:#### (2001)

Rational Ignorance at the Patent Office

the same effects as above (that is, a reduction in the number of patents issued and the number applied for, as well as a reduction in the number of
patents that survive to litigation), the cost-benefit structure is unchanged.
Another way to look at these numbers would be to say that society
was willing to commit $1.28 billion more than we currently do to weeding out bad patents. How should we spend that money? One way to
spend it would be to add eighteen hours to the time examiners spend on
each patent application (as noted above, the net cost of such a proposal is
$1.52 billion in increased prosecution costs minus $262 million in litigation and licensing savings, or $1.26 billion). Alternatively, we could
more than double the amount of effort put into determining validity in
court casesthat is, add over one thousand more hours of prior art
searches, expert analyses, and judicial hearings on each patent that is actually litigated. If we want to get the validity decisions right in the cases
that matter, the latter is surely a more efficient use of money than the
former.63
The strong implication of these numbers is that society ought to resign itself to the fact that bad patents will issue, and attempt to deal with
the problem ex post, when the patent is asserted in litigation. This result
is admittedly counterintuitive. It depends crucially on the fact that very
few patents are ever the subject of litigation, or even licensing. Because
of this, money spent improving the PTO examination procedures will
largely be wasted on examining the ninety-five percent of patents that
will either never be used, or will be used in circumstances that dont crucially rely on the determination of validity.64
My argument becomes more intuitive if we take the position of PTO
reformers a bit further. Suppose, for example, someone suggested that to
minimize the risk of error we should conduct the equivalent of a full trial
on validity (say, one thousand hours of examination) before granting a
patent. This would certainly reduce the risk of bad patents getting
through the system. But most people would rightly think such a suggestion ludicrous and unworkable. Why? Because they intuitively recognize that we simply cannot afford perfect decision making in each of the
hundreds of thousands of cases on which the PTO has to make decisions.
We understand rational ignorance on the part of the PTO, in other
wordsthe only question is how much time we should spend per patent.
From a cost perspective, the answer is not much.
rule, there is no duty to conduct a prior art search, and thus there is no duty to disclose art of which
an applicant could have been aware.). In part, too, companies fear that if they find a competitors
patent during a prior art search, they are on notice of that patent for purposes of willful infringement if they are later sued.
63
For this reason, it seems wrong to suggest that courts should defer to the PTO because of their
greater expertise in the technical fields in question; cf. Arti K. Rai, Addressing the Patent Gold
Rush: The Role of Deference to PTO Patent Denials, 2 WASH. U. J.L. & POLY 199, 217 (2000)
(suggesting such deference, albeit in the context of appeals from PTO rejections of patent applications). Whatever greater technical expertise does exist (and note that it is patent lawyers as well as
judges who contribute information to the litigation system) is counterbalanced by the greater time
devoted to validity in the litigation process.
64
Jay Thomas is one of the few to have recognized this point. He notes in passing that although Patent Office shortcomings are apparent, it is equally obvious that an exhaustive prior art
search for each application would be inefficient. Many patented inventions are never commercialized. Thomas, supra note 1, at 11. He does not pursue the implications of this idea, however.

15

NORTHWESTERN UNIVERSITY LAW REVIEW

IV. OBJECTIONS
Those who doubt this analysis might raise a number of objections.
Some might question the assumptions that are necessarily built into some
of these numbers. Others might question why I believe validity doesnt
matter in circumstances where the patent isnt litigated or licensed. Still
others might argue that the focus on legal costs ignores issues of deterrence, issues of delay and uncertainty, or issues of fairness and distribution. I consider each of these arguments in turn.
A. Varying My Assumptions
Perhaps you think Ive played fast and loose with my statisticsthat
my assumptions are unrealistic, and that a more realistic assessment of
the world would paint a very different picture. For that reason, this Subpart explores how some possible alternative numbers would affect my
analysis.
I start by noting that many of the numbers in this study reflect pretty
hard data. We know exactly how many patent applications are filed,
and how many patents issue every year. We know how many cases are
filed, and how many of them proceed to trial. We know how much patent litigation costs because we have survey data from the largest association of intellectual property lawyers in the country. Those numbers are
what they are. Other numbers, such as the costs of patent prosecution,
are estimates, but they seem to be reliable estimates. Not only do many
different sources concur, but they also seem to fit with both my personal
experience and the experience of those with whom Ive talked.
Where the numbers are less solid, notably in the cost and frequency
of licensing, I have tried to give PTO reformers the benefit of the doubt
by estimating conservatively (in this case, by estimating on the high side
both the amount of money spent on validity research in an average licensing transaction and the number of patents that are actually licensed
for a royalty). Nonetheless, you may not believe me. So lets change
some numbers. Suppose you think that fifteen percent of all patents are
licensed for a royalty: thats 22,500 different new patents licensed each
year.65 Further, lets say you think the average cost of the validity investigations done for a single patent in each of these 22,500 licensing
transactions is even higher than $50,000say $80,000 out of a total cost
per licensed patent of $130,000. Both of these numbers strike me as too
high, but never mind. Lets explore.
If fifteen percent of patents are licensed every year, thats 20,500
patents (once we subtract the 2,000 that go to litigation which we have
already accounted for). These 20,500 licensed patents, at $80,000 validity cost each, total $1.64 billion per year spent investigating validity;
20,500 licensed patents at $130,000 total licensing cost per patent,
amounts to $2.66 billion total spent on licensing per year. If we revise
the calculations in Part III, adding $2.66 billion to the $2.1 billion spent
on litigation, we have a total cost of $4.76 billion. If we divide this by
ten as we did in Part III, reflecting the reduced number of patents that
65

This represents far more than 22,500 new licensing agreements per year. A single revenueproducing patent may be (and often is) licensed to multiple entities over its lifetime.

16

95:#### (2001)

Rational Ignorance at the Patent Office

will be litigated and licensed, an enhanced examination program would


save us $476 millionstill far less than the $1.52 billion cost of adding
just a few hours to the examination of each application. Even at this unrealistic cost estimate of $130,000 per licensed patent, you would have to
believe that more than 80,000 new patents are licensed for royalties
every yearmore than half of the patents issued each year, and more
than pay their maintenance feesto believe that doubling the examination time would save as much money as it would cost. There may be uncertainty about how much licensing occurs, but there isnt that much
uncertainty.
The model is more sensitive to my assumptions about the number of
patents that would be weeded out by adding eighteen hours more examination time to each patent. Again, it seems to me that my assumptions
(that patent applications will actually go down as a result of this examination, despite their current trend, and that the new examination procedures will allow the PTO to weed out more than four times as many bad
patents than they currently do without affecting any of the good ones)
bend over backwards to be fair to PTO reformers. But lets relax those
assumptions anyway. If you believe that eighteen hours of additional examination time would allow the PTO to weed out all the patents that turn
out to be invalid in litigation (forty-six percent in recent years),66 and that
it would weed out a similar number of bad patents in licensing, the total
savings from the corresponding reduction in litigation and licensing
would be forty-six percent of the $2.62 billion currently spent on litigation and licensing, or $1.21 billion. That is, even if the number of litigated and licensed patents were cut almost in half by this simple change
in the PTO, it still wouldnt be cost-effective.
We could achieve a greater cost savings in litigation if we were to
eliminate validity litigation altogether, for example by conclusively presuming that a patent once issued was valid.67 This would save another
$625 million spent on validity investigations and litigation in cases in
which the patent really is objectively valid. With this new savings, the
reform proposal might in fact turn out to be cost-effective.68
Think for a moment about what this would mean, however. To costjustify doubling the examination time for all patents, we would have to
believe that the PTO would use the extra eighteen hours to make its examinations perfectthat is, that no bad patents would issue. And we
would have to be so sure of this judgment that we would abolish an accused infringers right to even try to persuade a court of the contrary, despite the fact that they have no right to participate in patent prosecution.
66

See infra note 128.


Cf. Kesan, supra note 1, at 26 (making a similar proposal contingent upon full disclosure of
prior art by the patentee). Kesans proposal is not for a conclusive presumption of validity, however,
but merely for an extremely high presumption. This more moderate proposal will actually save less
money than the proposal discussed in text.
68
Even this is not a sure thing, however. The patent laws were amended in 1999 to add a thirdparty reexamination process that competitors can use. At the moment, virtually no one uses this system, because anyone who does so is foreclosed from making validity arguments in litigation. 35
U.S.C. 311-318 (West Supp. 2000). If validity litigation were abolished, competitors would rush
to use this system as a second-best alternative. The costs of the system would have to be factored
into the analysis in the text.
67

17

NORTHWESTERN UNIVERSITY LAW REVIEW

Doing so would require substituting a very brief review (thirty-six hours)


of a patent for the thousands of hours and considered judgment of trial
and appellate courts. We should be willing to do this only if we truly believe that even brief consideration by a patent examiner is more likely to
discern the truth than the trial process. I suspect that no one really believes this is possible. And if you do in fact believe thirty-six hours is
enough for a perfect examination, its worth asking why litigants spend
thousands of hours on validity issues whenever a patent goes to court today, when they could presumably have an objectively correct answer for
a much lower price.69
In short, the central insight of this Essay is not an artifact of the assumptions Ive made. Even if we relax those assumptions beyond all
reasonable bounds, the fundamental fact remains that litigation of a few
patents is a far more efficient way of determining validity than giving a
detailed ex ante examination to all patents.
B. Does Validity Matter for Unasserted Patents?
There is one assumption I have made that we have not yet considered. That is the assumption that the objective validity of a patentthe
right answer, if you will, to the question of whether the patent should
ever have issuedshould matter only in cases in which the patent is asserted against an accused infringer, either in litigation or in licensing.
This also might strike some people as counterintuitive. Wouldnt venture capitalists, competitors, and the company itself care whether the patent is a good one or a bad one, even if they dont themselves intend
to pay money for it? In fact, though, Im not sure thats true. To start
with the easy case, significantly more than half of all issued patents are
never used for any purpose whatsoever, except maybe to make their
owners or inventors feel good about themselves.70 Surely it doesnt matter much whether these patents are really valid in any objective sense.
These patents dont impose any direct cost on anyone, though there
might be some indirect clutter cost associated with issuing so many
patents.71
Patents which are put to some use other than licensingimpressing
investors, adding to asset counts, helping to define a market strategy, advertising ones patented technology or marking products, and so on
also dont seem to me to require objective validity. It is true that the existence of the patent itself is being used for some purposehere, generally a communicative one. But the harm caused by letting even bad
69
Indeed, some evidence of how the patent bar views the PTO examination system is the fact
that essentially no one has opted for the new inter partes reexamination system over court litigation.
See supra note 68. The inter partes system allows competitors to participate in reexamination and
bring their prior art to the attention of patent examiners. If it were reasonable to believe that the
PTO would do a better job of examining patents than the court system, one would expect that many
competitors would use this system to reevaluate the validity of patents in lieu of litigation. So it is
striking that competitors have not only voted with their feet for litigation over PTO examination, but
have done so unanimously.
70
This percentage includes not only those patents for which maintenance fees are not paid, but
also some patents for which fees are paid, but which are still never used.
71
This cost is likely to be minimal, if indeed it exists at all. The only reason competitors search
the patent database is if they are or are likely to be involved in litigation.

18

95:#### (2001)

Rational Ignorance at the Patent Office

patents be used for such a purpose seems to be minimal. Sophisticated


parties like venture capitalists or a potential merger partner generally understand what patents are, and what it means to have one. They are capable of discounting for the risk that the patent would be held invalid if it
were ever enforced.72 While those parties might benefit from knowing
the patents in question were objectively valid, there is no proposed system that will provide such information. The question is whether there is
a significant benefit to society from increasing the probability that those
communicative patents are valid; that would be true only if those who
rely on information about these patents couldnt assess the probabilities
easily.
Cross-licensing of multiple patents without royalty payments presents the strongest intuitive case for caring about the validity of a patent.
Why should a party want to take a license to a patent that isnt really
valid anyway? One answer is the one just given: parties in crosslicensing transactions tend to be sophisticated entities with patents of
their own, who understand and can discount for the possibility that the
patent might be objectively invalid. Further, these licenses tend to involve multiple patents on each side, and (in the circumstances I have
postulated) the parties essentially trade patent rights, rather than one
party paying anything to the other. In these circumstances, it hardly
seems likely that the objective invalidity of one patent in this group
would kill the deal, though it conceivably could affect the decision to
make the cross-license royalty-free in marginal cases.
C. The Social Cost of Bad Patents
Probably the most common objection to the claims made in Part III
will be that they ignore the social cost of issuing bad patents. This is not
entirely true. The social cost of bad patents comes in several forms.
First, bad patents that are litigated impose litigation costs on society;
those costs have already been considered.73 Second, bad patents that are
licensed impose legal fees on licensees; those costs too have already
been considered. Third, some licensees may pay a royalty rather than
fight even a bad patent in court. Those royalty payments are a social cost
to bad patents that I have not yet considered. Finally, it is possible that
the mere existence of bad patents that arent litigated or licensed may
nonetheless deter some lawful competitive conduct. In this Subpart, I
72
Indeed, venture capitalists often evaluate companies on the basis of patent applications,
where the determination necessarily involves uncertainty even as to whether the patent will issue at
all.
Because venture capitalists can be expected to discount the value of patents, the overinvestment critique offered by Lunney and others doesnt apply here. See Glynn S. Lunney, Jr., Reexamining Copyrights Incentives-Access Paradigm, 49 VAND. L. REV. 483 (1996) (arguing that overly
strong intellectual property rights promote overinvestment in intellectual property, inefficiently
drawing assets away from other productive sectors of the economy).
73
I have not included patent damages granted at trial as a social cost of bad patents. We do not
have a perfect measure of which patents are objectively valid and which objectively invalid, but it
seems unreasonable to assume both that the courts will erroneously validate bad patents, and that
some marginal improvements in the patent system would have weeded out those same patents. So if
a patent is held valid and infringed in court, the damages that ensue are a cost to the patent system as
a whole, but they arent attributable to bad patents.

19

NORTHWESTERN UNIVERSITY LAW REVIEW

consider these latter two forms of social cost to see if they change my
analysis.
The social cost of issuing bad patents is different than the social cost
of the patent system itself. The patent system intentionally restricts
competition in certain technologies to encourage innovation. Doing so
imposes a social cost, though the judgment of the patent system is that
this cost is outweighed by the benefit to innovation.74 That social cost is
not at issue here; rather, it is only the cost of issuing marginal bad patents
that matters for my analysis.
1. In Terrorem Effects.A more serious concern is that potential
competitors or follow-on innovators in a field might be deterred from entering the field by the existence of patents owned by their competitors.
On this view, when a patentee obtains a patent, the existence of the patent itself sends a powerful signal to competitors: stay away. If patents
do indeed have such an inherent in terrorem effect, then issuing bad
patents has a real cost to society. The mere fact that those bad patents issued will deter some socially optimal behavior by competitors, imposing
a cost on society that I have not yet taken into account.
There is something to this objection. Certainly the issuance of bad
patents has the potential to deter competition that should be lawful in
some marginal cases. But this concern can be overstated. First, by hypothesis we are talking only about the subset of patents that are never
litigated or licensed. The vast majority of these patents simply exist; the
in terrorem concern requires us to believe that competitors are regularly
searching patent databases to make sure they are not infringing a patent
that no one has brought to their attention. In my experience, this is simply not the way businesses operate. If they think about the problem at
all, most sophisticated companies strenuously avoid reading other companies patents because they dont want to be charged with knowledge of
the patents existence.75 But far more companies simply dont consider
their potential patent liability unless or until a patent is brought to their
attention.
So the patents with the most likely in terrorem effect must be ones
that the patentee has tried and failed to license. In those cases, competitors or potential competitors may have received a letter alerting them to
the existence of the patent, though that letter will likely have been
74
There is a great deal of literature attempting to assess whether that judgment is accurate or
not, usually without success. George Priest complained years ago that there was virtually no useful
economic evidence addressing the impact of intellectual property. George Priest, What Economists
Can Tell Lawyers About Intellectual Property, 8 RES. L. & ECON. 19 (1986). Fritz Machlup told
Congress that economists had essentially no useful conclusions to draw on the nature of the patent
system. See An Economic Review of the Patent System Before the S. Subcomm. on Patents, Trademarks, and Copyrights of the S. Comm. on the Judiciary, 85th Cong. 55 (1958) (statement of Fritz
Machlup). For some of the disagreements among historians over the impact of the patent system on
innovation, see ROBERT P. MERGES ET AL., INTELLECTUAL PROPERTY IN THE NEW TECHNOLOGICAL
AGE 126-27 (2d ed. 2000). That broader analysis is beyond the scope of this Essay; suffice it to say
that if the patent system as a whole is a bad idea, as some claim, neither the proposed reforms I discuss nor my objections to them matter very much.
75
Continuing to infringe a patent once you become aware of it makes you liable for willful infringement, which can justify an award of treble damages against you. See, e.g., Comark Comm.
Inc. v. Harris Corp., 156 F.3d 1182 (Fed. Cir. 1998).

20

95:#### (2001)

Rational Ignorance at the Patent Office

couched in extremely bland terms, purporting merely to make the competitor aware of the patent in case it should want to take a license.76 The
recipient may ignore the letter,77 or it may react. The most common
reaction is to obtain an opinion letter from counsel advising whether or
not the patent is valid, and whether or not the recipient infringes.78
Based on the advice of counsel, the company might conceivably change
its business plans, but the recipients counsel will rarely advise such a
course of action when the patent is objectively invalid. Another possible
reaction is to give in without a fight. Some companies, especially small
ones unsophisticated in patent law, may simply drop their plans to sell a
product once a patent is brought to their attention, without even entering
into a license negotiation. Again, though, this is an unusual reaction.
I dont mean by the preceding discussion to dismiss the in terrorem
effect altogether. But I do think it is generally strongest in cases in
which the patent is litigated or licensed at some point in its life, and in
cases in which the patent is objectively valid. Once we exclude those
cases, as we must do in measuring the merits of my thesis, the effect is
likely to be a small one. Nonetheless, in part because it is hard to quantify any chill to innovation that might result, I have taken this concern
into account in my suggestions for reform.79
2. Holdup Licensing.Patent owners might try to game the system by seeking to license even clearly bad patents for royalty payments
small enough that licensees decide it is not worth going to court. This
sort of nuisance value claim occurs in all sorts of contexts, most notably in securities litigation.80 When it occurs in the patent system, as it
doubtless does in some cases, the result is inefficiency in societys allocation of resources. Innocent competitors that are not infringing a valid
patent nonetheless pay money to the owners of invalid patents; this transfer encourages patenting and discourages competition to a greater extent
than is socially optimal.
It is hard to know how much money is transferred from licensees to
76
The letters are written in this roundabout way because if the patentee actually comes out and
suggests that the recipient infringes the patent, the recipient has standing to file suit for declaratory
judgment of noninfringement, an act which confers significant procedural advantages. See Amana
Refrigeration, Inc. v. Quadlux, Inc., 172 F.3d 852 (Fed. Cir. 1999) (holding that under the Declaratory Judgment Act, a district court has jurisdiction if there is an actual controversy, and that an actual
controversy exists if there is an explicit threat by the patentee which creates a reasonable apprehension on the part of the declaratory plaintiff that it will face an infringement suit, and present activity
which could constitute infringement); Kimberly A. Moore, Forum Shopping in Patent Cases: Does
Geographic Choice Affect Innovation? 79 N.C. L. REV. 889, 920 (2001). As a result, patent lawyers
invariably advise their clients to write neutral letters that dont give the recipient a reasonable apprehension of being sued.
77
One general counsel told me that his company never does anything about such a letter unless
it gets a follow-up letter or call from the patentee.
78
As a result, one could get a sense of the dimensions of this problem by collecting data on how
often companies get opinions from patent counsel involving patents that never end up being litigated
or licensed to anyone. I suspect this is a relatively infrequent occurrence, but I have no data to support or refute that hunch.
79
See infra notes 99137 and accompanying text.
80
Cf. Janet Cooper Alexander, Do the Merits Matter? A Study of Settlements in Securities Class
Actions, 43 STAN. L. REV. 497 (1991) (examining securities cases that settled for nuisance value regardless of merit).

21

NORTHWESTERN UNIVERSITY LAW REVIEW

patentees through such holdup claims, and (more important for my purposes) how much less money would be transferred under a more robust
examination system. But with some plausible assumptions, we can generate some ballpark figures. To calculate an upper bound on the avoidable social costs from holdup licensing, I begin by recognizing that the
possible number of holdup situations is limited by the total number of
patents licensed for revenue or litigated but settled. There are 6,750 such
patents per year 5,250 licensed without litigation, and 1,500 litigated
but not tried. Next, the avoidable holdup situations are a function of how
many of those patents will be weeded out through eighteen hours additional examination.81 I earlier determined a high estimate for that number to be around ten percent of total patents. This means that, at most,
675 new patents per year are involved in avoidable holdups. This number is no doubt an exaggeration because it assumes that every instance of
licensing without litigation is in fact a holdup attributable to a bad patent
and not to a legitimate business deal. A more plausible (but still extremely high) number would be that half of those deals are legitimate,
and the other half are holdups. If so, we are talking about 338 patents
involved in holdups. Those patents may be asserted against multiple parties.
What is the cost to industry of such holdups? Presumably each licensee being held up would not be willing to pay more than the cost of
litigating the patent. If the patentee demands more, the licensee ought to
be willing to go to trial instead.82 The weighted cost of litigation is
$656,000.83 The total cost depends on how many defendants will be held
up in this way. Earlier, we assumed that the average number was two per
patent, a blended estimate reflecting the number of potential licensees actually approached, potential cost savings if defendants pool their resources, and the increased likelihood that at least one case will result in
litigation as the number of potential licensees increases. Thus, even under the assumptions made here the maximum likely social cost of licensing holdups is 338 patents times two licensees each times $656,000, or
$443 million.84 This is certainly a cost to take into account, but it doesnt
81
The discussion in text likely overstates the extent to which the holdup problem can be
avoided. If an applicant is serious about asserting its patent, it is unlikely to give up easily in prosecution. Since it is impossible under current PTO rules for an examiner to ever finally reject an application, adding examination time to such a patent may simply delay the process and increase its
cost without changing the ultimate outcome.
82
To be sure, this assumes the licensee can determine relatively easily that the patent is invalid.
If that is not true, perhaps because of some information asymmetry, the licensee should pay the expected cost of litigation plus the expected payout (the likely damages times the probability of losing
the case). But in this particular context, it seems reasonable to assume that the licensee will figure
out pretty easily that it is dealing with a bad patent. After all, the only cases we are considering here
are cases that patent examiners will determine to be invalid based merely on eighteen hours more
examination. These are presumably not the close cases that might ordinarily create uncertainty
about trial outcomes.
83
This is based on the calculations supra notes 28-32 and accompanying text. In brief, one case
in sixteen costs $1.503 million to litigate, while under the assumptions made there the remaining fifteen cases cost $599,000 on average to litigate.
84
In economic terms, the social cost of such a holdup is not the price actually paid by the licensees, but the deadweight loss caused by the resulting increase in licensee costs. That deadweight
loss is in turn a function of the percentage increase in cost resulting from the license payment and the
shape and elasticity of the demand curve. Even if the license payment were equivalent to a monop-

22

95:#### (2001)

Rational Ignorance at the Patent Office

outweigh the increased social cost of strengthening the prosecution system.


3. Treating Patents in Isolation.A final limitation on the social
cost of in terrorem and holdup effects involves the practical context in
which most patents are actually asserted or licensed. While some patents
are licensed in a stand-alone transaction, most patents are licensed either
as part of a group of related patents, or as part of a hybrid transaction
that bundles patents with trade secrets and know-how.85 In these cases, it
may be difficult to separate out a licensees motivation for entering into a
license agreement. Some of the social costs estimated in this section
may be overstated, therefore, because a license transaction would have
occurred whether or not a patent that was only one part of a larger deal
was objectively invalid.
4. Facilitating Collusion.Licensees might agree to pay royalties
on patents they know are invalid as part of a scheme to cartelize an industry. Thus, one potential cost of bad patents could be a cost to competition in highly concentrated industries. This is a real risk, though it is an
easy one to overstate. In most industries in which cartels are feasible,
they will form around a variety of mechanisms. It seems a stretch to say
that the grant of a bad patent will enable a successful cartel that otherwise would never have come into being.
Nonetheless, the law should be vigilant to prevent such abuses of
patent rights. Antitrust law forbids the use of patent licenses as a tool to
cover a cartel.86 The Courts decision in Lear, Inc. v. Adkins87 may also
provide incentives for licensees to challenge the validity of patents used
to facilitate a cartel. Lear has come under sustained attack of late by
both courts and commentators,88 but the fear of collusive licensing may
actually be a good reason to retain Lear and permit licensees to challenge
a patents validity at any time.
D. The Costs of Delay and Uncertainty
A more serious objection to my model is that it considers only the
costs of prosecution and litigation, and does not consider the costs imoly price a decidedly unlikely assumption for a nuisance-value payment the deadweight loss
would not exceed the total of the license payment under normal circumstances. It could never do so
if the demand curve were linear. Even under a hyperbolic demand curve, the deadweight loss would
exceed the license payment only if the amount produced under competition (QC) was nearly three
times the amount produced under monopoly (QM). More precisely, QC > 2.71QM. [Proof omitted].
While such circumstances are theoretically possible, they are unlikely. As a result, I use the total
amount of the license payment as an estimated upper bound of the social cost from holdups.
85
See Fish & Richardson P.C., Trade Secrets/Know-How and Hybrid Licenses, available at
http://www.fr.com/publis/f_paper10.html ([T]he trend is towards license packages that include a
mix of rights.).
86
15 U.S.C. 1. See generally HERBERT HOVENKAMP ET AL., INTELLECTUAL PROPERTY AND
ANTITRUST LAW ch. 34(B) (forthcoming 2001).
87
395 U.S. 653 (1969).
88
See, e.g., Studiengesellschaft Kohle v. Shell Oil Co., 112 F.3d 1561, 1567 (Fed. Cir. 1997)
(noting the Federal Circuits attempts to distinguish later cases from Lear, denying protection of the
Lear doctrine); Rochelle Cooper Dreyfuss, Dethroning Lear: Licensee Estoppel and the Incentive to
Innovate, 72 VA. L. REV. 677, 680 (1986) (arguing that a major flaw in the Courts analysis was its
failure to consider the economic function played by licensee estoppel).

23

NORTHWESTERN UNIVERSITY LAW REVIEW

posed by a delayed resolution of the validity question. Having to go


through litigation to determine validity not only costs a great deal of
money, but it takes quite a bit of time. Indeed, the average time between
the issuance of a patent that would later be litigated and a final decision
on its validity in litigation was 8.6 years. For many patents, the validity
decision was not made until thirteen or fourteen years after the patent issued.89 During this period, both the patentee and potential infringers (accused and unaccused) are uncertain about their legal rights. Surely we
would be better off knowing sooner rather than later whether a patent is
valid.90
This is a valid objection. All other things being equal, we would
like to resolve the validity of patents as early as possible. But all other
things are not equal. For one thing, unless we are willing to give up validity litigation altogether and rely only on the judgment of the PTO as to
validity, issued patents will still be subject to validity challenges, and
thus the validity of litigated patents will continue to be uncertain ex ante
even under an enhanced examination system. The only increase in certainty that such a system would achieve would come in those cases in
which an enhanced examination system would reject a patent application
that is allowed under the current system. For this subset of patents, generously estimated at ten percent of the total in Part III, there would be no
patent under the new system, hence no litigation at all and therefore no
uncertainty. But we are talking at most about only a couple hundred patents a year in this situation.91
Further, delay and uncertainty are not only factors in litigation.
They affect patent prosecution too. An enhanced examination system
will necessarily increase the delay in issuing patents, and therefore the
uncertainty associated with the ownership of legal rights in an invention.
This delay will primarily affect patent applicants, who will have to wait
longer to learn whether they will get a patent and, if so, must wait longer
to assert it.92 This delay wont be as long in any individual case as the
delay caused by litigation. But it will affect all 275,000 applications
every year.
Increasing the examination scrutiny imposed on patents will also
likely result in the rejection of some objectively valid patents. In the discussion in Part III, I assumed that more examination time would enable
89

See Allison & Lemley, supra note 4, at 236 tbl.12.


Thus, Craig Nard argues in favor of a post-grant opposition proceeding on the grounds that it
will increase certainty ex ante. See Craig Allen Nard, Certainty, Fence Building and the Useful Arts,
74 IND. L.J. 759, 795 (1999). But see Ian Ayres & Paul Klemperer, Limiting Patentees Market
Power Without Reducing Innovation Incentives: The Perverse Benefits of Uncertainty and NonInjunctive Remedies, 97 MICH. L. REV. 985 (1999) (arguing in favor of uncertainty as to the validity
of a patent).
91
Weeding out bad patents should increase the probability that those patents that do make it
through are valid. This in turn might affect decisions to litigate. But it would do so in unpredictable
ways. For example, an increased probability of validity might induce more defendants to settle
without litigation or at least before trial. On the other hand, the same increased probability might
induce more patentees to litigate or license their patents. The net effect of these effects is unclear.
92
Prosecution uncertainty may also affect potential infringers, since under the 1999 changes to
the U.S. patent system most patent applications will be published eighteen months after they are
filed. See 35 U.S.C. 122(b) (West Supp. 2000). Potential infringers who are aware of those published applications will have to wait longer to see whether a patent issues on the application.
90

24

95:#### (2001)

Rational Ignorance at the Patent Office

the PTO to weed out more bad patents without weeding out any good
patents. But that is unrealistic. Systems that operate under uncertainty
always balance type I and type II errorsfalse positives (here issuing
bad patents) and false negatives (here failing to issue good patents). If
we fail to issue good patents, we may reduce innovation incentives and
interfere with the other socially valuable uses of patents, such as their financing function.
In short, delay and uncertainty impose costs on both sides of the
equation. Balancing these effects is difficult, and I dont propose to do it
here. Rather, I merely suggest that the costs of delay and uncertainty resulting from any given policy be taken into account in deciding what to
do. I return to this issue in Part V.
E. Distributional Effects
There is one fundamental difference between the social costs of patent prosecution and the social costs of licensing and litigation. The former costs are internalized by patent applicants, who not only pay their
own lawyers but also pay (through PTO fees) the costs of examination.93
By contrast, the costs of licensing and litigation are imposed not just on
patent owners, but on accused infringers and, indirectly, on the public.94
Thus, relying more heavily on litigation rather than prosecution to determine patent validity has distributional as well as efficiency consequences.
This, too, is a fair concern. However, it is subject to some of the
same caveats as the prior objection. First, enhanced examination will not
eliminate validity disputes in litigation or licensing in the overwhelming
majority of cases. These activities by their very nature impose costs on
third parties; it is only in the relatively small set of cases that would
never be brought (because the patents that are their subjects would never
have issued) that any savings in cost will occur.
Second, it seems to me a contestable proposition that all the costs of
determining validity should be borne by the patent applicant. It is true
that the applicant is the one who would benefit from the issuance of the
patent, so it might seem fair that the applicant should bear all the cost.
But our society issues patents not because it desires private monopolies,
but because granting patents in appropriate circumstances serves the public interest. It is equally plausible to argue that accused infringerswho
have the most direct interest in the outcome of the validity proceeding
and taxpayers should help shoulder the burden of determining who is
right in this area, just as they do in any other sort of lawsuit.
Finally, this sort of cost-sharing serves instrumental goals as well.
One problem with resting the validity decision entirely on examiners is
93

Indeed, they currently pay more than this, because of Congresss unconscionable decisions
over the last several years to take money from the PTO user fees accounts and use them for the general federal revenue. See, e.g., Lisa Seachrist, House Committee Proposes Taking $295M of Patent
Fees, BIOWORLD TODAY, June 22, 2000 (discussing fee diversion).
94
I should make it clear that I have not considered as a cost of the patent system any royalty
fees or damages paid by defendants as a result of infringing objectively valid patents. Not only are
these costs a pure transfer from one party to another, they are a transfer explicitly intended by the
patent statute.

25

NORTHWESTERN UNIVERSITY LAW REVIEW

that, no matter how well qualified they are, they have no direct interest in
the outcome. This is a good attribute in a decision-maker, but not such a
good thing in a party advocate. Our system of litigation is founded on
the idea that those with the most interest in an outcome will do the best
research and make the best arguments for their position, and the truth
will win out. On this view, the fact that accused infringers have to pay
some of the cost of determining validity is not a bug in the system, but a
feature. They will do a better job of proving a patent invalid than an
examiner ever could.
Having said all this, I agree that one factor to consider in deciding
how to allocate costs is that determining validity ex ante imposes fewer
costs on third parties than waiting to determine validity ex post. I discuss
some implications of this in Part V.
F. The Credibility of the Patent System
Finally, one might object that issuing bad patents brings the system
into disrepute, which in turn may reduce the value of good patents, and
therefore harm innovation. It is certainly true that people will think less
highly of patents as a class if they know that many of those patents are
objectively invalid. But I think this concern is overstated, for several
reasons. First, for many of the patents I have discussed, especially the
trophy patents and those that sit on the shelf, it is not clear that the level
of quality of the patent system is all that important. Second, innovators
adapt to the legal background they are given. There is no examination of
trade secrets at all, essentially no examination of copyrights, and no required examination of trademarks. Yet in each case society builds assumptions (and indeed transactions) around the existence of these
inchoate rights. Its even easier to obtain a bad trade secret or copyright claim than a bad patent, but that doesnt prevent trade secrets and
copyrights from having value. As suggested above, venture capitalists
and licensees already make the same calculation for patents.95 This
brings me to my final point: it is implausible to think that any examination system could ever eliminate uncertainty over the validity of a patent.
So those involved with the patent system will always have to make this
calculation.
Two things about credibility do seem important. First, the legal
rules should accurately reflect reality. If the PTO issues lots of bad patents, the legal system should take that fact into account in deciding the
power of patents. I suggest below some ways to bring the law more in
line with reality, notably by scaling back the presumption of validity.96
Second, patentees may want some signaling mechanism by which they
can distinguish good patents from bad (or at least run-of-the-mill) ones.
A Japanese-style elected examination system may allow such sorting.97
V. IMPLICATIONS FOR OTHER REFORM PROPOSALS
So far I have focused attention on suggestions that the PTO spend
95
96
97

26

See supra notes 7072 and accompanying text.


See infra notes 119132 and accompanying text.
I discuss the Japanese system infra notes 117118 and accompanying text.

95:#### (2001)

Rational Ignorance at the Patent Office

more time examining patents to weed out bad ones. But several other
proposals for PTO reform have been floating around. In this Part, I consider the implications of my analysis for those proposals.
A. Improving PTO Processes
Nothing in my analysis suggests the PTO should not spend the
money it is allocated as efficiently as possible. Some recent reforms
within the organization have been aimed at enabling the PTO to operate
more efficiently;98 I see no problem with this. Similarly, Rob Merges
makes a series of suggestions for improving the incentives for examiners.
For example, Merges suggests changing the existing examiner incentive
system so it rewards not only those who grant patents, but also those who
reject applications in appropriate circumstances.99 He also suggests better training of examiners.100 These suggestions are perfectly consistent
with my analysis, though one should be leery of changes that involve a
significant increase in total PTO costs.101
Similarly, my analysis does not militate against increasing the
budget of the patent office for reasons other than enhancing the examination each patent receives. For example, some people have suggested that
the PTO needs to hire more examiners so that it may make decisions
more quickly, holding the level of examination constant.102 This may or
may not be true, but it is irrelevant for my analysis. Reducing the time
patents spend in prosecution would presumably benefit patent holders in
general, not just the subset who litigate their patents.
Finally, reform at the PTO might take the form not of changing its
level of effort, but of changing the legal standard it is bound to apply.
The current legal standard assumes an application is patentable; the burden is on the examiner to present a prima facie case of invalidity.103
There is a plausible legal argument for reversing this presumption and
requiring patent applicants to shoulder the burden of proving their invention is patentable. Whether reform along these lines would increase the
cost of the system is a complex question. While applying the new standard in the first instance would probably not be any more costly, the
unique nature of the U.S. patent prosecution system, in which it is virtually impossible for an examiner to reject a patent application for good,104
98

The American Inventor Protection Act of 1999 established the PTO as a quasi-independent
government corporation within the Department of Commerce, giving it more administrative
autonomy. See generally 35 U.S.C. 1(a) (West Supp. 2000) (laying down the history, authority,
and responsibilities of the PTO).
99
Merges, supra note 1, at 606-09. The Quillen-Webster study dramatically demonstrates the
need for such a reform. See Quillen & Webster, supra note 11.
100
Merges, supra note 1, at 607.
101
Thus, Merges also suggests paying examiners salaries competitive with what engineers earn
in the private sector, which might involve doubling the PTOs labor budget. See id. at 606-607 (noting representative figures for salaries).
102
I confess that Ive argued this myself. See Lemley, supra note 22, at 147. Congress was sufficiently concerned about the possibility of delay caused by the PTO that it amended the patent statute in 1999 to insulate patent applicants from loss occasioned by examiner delay. See 35 U.S.C.
154(b).
103
See, e.g., In re Kotzab, 217 F.3d 1365, 1371 (Fed. Cir. 2000).
104
See, e.g., Quillen & Webster, supra note 11 (discussing the process of abandonment and refiling).

27

NORTHWESTERN UNIVERSITY LAW REVIEW

may mean that a more stringent examination standard will simply result
in a more protracted prosecution process.
B. Third-Party Oppositions
A number of commentators have suggested that the United States
should adopt a pre-grant opposition system like those in Europe and Japan, under which interested third parties could challenge applications before the PTO issues the patents.105 Historically, the U.S. has not allowed
third parties to participate at all in examining patents, and has given them
only a limited role in calling for an ex parte reexamination of a patent
after it has issued. Recent changes to the law would permit third party
participation in reexamination, but these changes still dont permit third
parties to oppose an application before it issues.106 Further, the new
third-party reexamination system is unlikely to be used much, because
anyone who participates in this system must forego their right to challenge validity in court if they are ever sued for infringement on the patent.107
Whether an opposition system is consistent with my analysis is a
difficult question. On the one hand, a pre-grant opposition proceeding
would significantly increase the cost of prosecution in cases in which it is
used.108 Thus, one might class it as merely a variant of the enhanced
prosecution system. But there is a significant difference between enhanced prosecution and third-party opposition. Oppositions will only be
filed in certain cases, and presumably a minority of cases. Merges reports that in Europe between 1994 and 1997, only about 6.5% of issued
patents were opposed.109 If the patent applications that are opposed are
also the ones most likely to be licensed or litigated, the opposition
proceeding begins to look more like an earlier and cheaper type of
litigation proceeding. F.M. Scherer has offered some evidence that this
is in fact the case.110
105

See, e.g., Dreyfuss, supra note 88; Mark D. Janis, Rethinking Reexamination: Toward a Viable Administrative Revocation System for U.S. Patent Law, 11 HARV. J.L. & TECH. 1 (1997);
Merges, supra note 1, at 610-15; Nard, supra note 90, at 795; J.H. Reichman, From Free Riders to
Fair Followers: Global Competition Under the TRIPS Agreement, 29 N.Y.U. J. INTL L. & POL. 11,
31 (1997). Jay Thomas has proposed an interesting variant on pre-grant opposition, in which parties
who submit patent-defeating prior art are paid a bonus or finders fee. See Thomas, supra note 1.
106
See 35 U.S.C. 122(c) (precluding pre-grant opposition to published patent applications).
107
35 U.S.C. 315(c). Thus, commentators (and everyone in the patent bar Ive talked to on
the subject) think the third-party examination system as currently written is essentially worthless.
See, e.g., Janis, supra note 17, at 481; see also Thomas, supra note 1, at 26 (describing reexamination as one of the greatest failures of the modern patent system). For an explanation that identifies
some limited circumstances in which the system might be used, see M. Patricia Thayer et al., Inter
Partes Reexamination: The United States Joins Europe and Japan in Providing an Adversarial Administrative Procedure for Testing Patent Validity (2001) (working paper, on file with author).
108
An opposition system would also introduce some of the elements of the litigation system into
prosecution: delay, uncertainty, and the imposition of costs on third parties (though in this case only
third parties that had voluntarily taken on the obligation). The amount of cost and delay would of
course depend on the nature of the system. But one probable parallel is interference proceedings, a
feature of existing patent law under which two applicants for a patent on the same invention dispute
priority. Despite the fact that only priority of invention is at stake, this inter partes proceeding is notoriously protractedsome cases have taken close to 30 years to resolve. See, e.g., Standard Oil Co.
(Indiana) v. Montedison S.p.A., 664 F.2d 356 (3d Cir. 1981) (resolving an interference proceeding
based on 1953 patent applications).
109
Merges, supra note 1, at 613 tbl.2.

28

95:#### (2001)

Rational Ignorance at the Patent Office

the case.110
Thus, an opposition system might be consistent with the insight of
this article, if one believes that applications that are opposed tend to be
the ones that are later litigated (or at least licensed). But if this is not
trueif there is a substantial disconnect between applications that would
be opposed and patents that are ultimately litigatedan opposition system could be a waste of time and money on everyones part.111
C. A Registration System
Some commentators have taken the opposite tack to the one criticized here, suggesting that the examination system should be thrown out
altogether and replaced with some sort of a registration or petty-patent
system. Most of these suggestions are variants on a registration system,
rather than a pure registration system, because the registration process
would coexist with the existing patent structure.112 But the basic argument is an extreme form of the one I have made here: that not only
shouldnt we increase the time spent in prosecution, we should reduce or
eliminate examination entirely, and rely on the litigation process to sort
the good patents from the bad.
This argument is not as radical as it sounds. Patents are the only
part of intellectual property law that actually requires government examination and approval as a prerequisite to filing a lawsuit. In copyright
and trade secret law, the government doesnt need to issue a copyright
or trade secret for the owner to go to court.113 While there is an examination system in trademark law, trademark owners can file suit even if they
dont register their marks.114
110
See Dietmar Harhoff et al., Citations, Family Size, Opposition and the Value of Patent Rights
(1999) (working paper, on file with author) (providing a study of German oppositions as a predictor
of patent value, finding that an opposition proceeding is the single best predictor of the ultimate
value of a patent).
111
The metric used to analyze Thomass proposal for bounties paid to those who submit invalidating prior art should be similar. If there is reason to believe that the bounties would encourage
prior art submissions primarily about patents that are likely candidates for future litigation, it may
serve a useful purpose. However, if the bounties encourage prior art submitters to pick lowhanging fruit by submitting art to invalidate obviously worthless patents, they may increase the cost
of the system with little corresponding benefit. Unless great care is taken in designing the system, I
fear the latter is the more likely result.
112
Thus, Ann Bartow calls for an origination patent system to provide a lesser form of protection to those who mostly want patent protection as a marketing device. See Bartow, supra note 47,
at 16-25. Richard Stern promotes a petty-patent system as a hybrid between patent and copyright
protection that might be appropriate for software. See Richard H. Stern, On Defining the Concept of
Infringement of Intellectual Property Rights in Algorithms and Other Abstract Computer-Related
Ideas, 23 AIPLA Q.J. 401 (1995). But see Mark D. Janis, Second-Tier Patent Protection, 40 HARV.
INTL L.J. 151 (1999) (being more critical of such second-tier patent systems).
113
There is a copyright registration system, and applying for registration is a prerequisite for a
United States copyright owner to file a lawsuit. Tellingly, however, even in those rare cases in
which copyright registration is refused, the refusal itself is sufficient for the copyright owner to go
ahead with the infringement lawsuit. 17 U.S.C. 411(a) (West Supp. 2000).
114
See 15 U.S.C. 1125(a). There are advantages to trademark registration, of course. Registration gives constructive nationwide use and priority, 15 U.S.C. 1057(c), 1072; it permits a mark
to achieve incontestable status, 15 U.S.C. 1065; it gives the mark owner the right to invoke the
assistance of customs officials in barring importation, 15 U.S.C. 1124; and it gives the mark owner
enhanced remedies, 15 U.S.C. 1116-1120.

29

NORTHWESTERN UNIVERSITY LAW REVIEW

My analysis may give some comfort to proponents of a registration


system. After all, if it would cost much more to expand prosecution than
to litigate the issue of validity after the fact, maybe the contrapositive is
also true: it may be more cost-effective to reduce the time and money
spent on examination below its current level, even at the cost of increasing the number of lawsuits and licensing deals.
Nonetheless, I do not endorse a pure registration system, for several
reasons. First, a true registration system would mean more than merely
reducing the number of hours devoted to examinationit would mean
eliminating examination altogether. Even if a reduction in prosecution
expenditures would be cost-effective, it does not follow that eliminating
them would be. At the very least, we would want more concrete estimates than we currently have of how many people would register and enforce patents under such a system and how many of the additional
applications would be spurious. Second, while I am not fully persuaded
by the arguments noted in the previous sections regarding in terrorem deterrence, holdup, delay, uncertainty and cost shifting,115 I do think they
have some force. Beyond a certain point, it may be unwise to shift the
burden of determining validity so far away from the patent applicant and
towards its competitors. Finally, examination may serve a socially useful
function by requiring patentees to restrict the scope of their claims, and
therefore help prevent some abusive claims that cover entire industries.116
A possible compromise solution is a hybrid registration-examination
system along the lines of the Japanese system.117 Under a hybrid system,
a patentee would register its invention, but the invention would not be
examined unless the patentee elected examination. A patentee could not
enforce the patent against a third party unless it first had the patent examined.118 The virtue of this system is that it forces the patentee to disclose
its private information about the quality and likely use of its patent by
deciding whether and when to have the patent examined. Because only a
subset of patents would be examined, and that subset are presumably
considered valuable by patentees and are likely candidates for litigation,
the examiners in a hybrid system could afford to devote substantially
more resources to those patents. Such a system therefore has the virtue
of more cost-effectively improving the examination process than the proposals I criticize.

115

See supra notes 7380 and accompanying text.


This is particularly true in light of the Federal Circuits en banc decision in Festo Corp. v.
Shoketsu Kinzoki Kogyo Kabushiki Co., 234 F.3d 558 (Fed. Cir. 2000) (en banc), cert. granted, 69
U.S.L.W. 3673 (2001), precluding resort to the doctrine of equivalents once the patentee has
amended a claim element during prosecution.
117
See T. TANABE & HAROLD WEGNER, JAPANESE PATENT LAW (1979) (describing the Japanese patent system).
118
This feature is particularly important, and distinguishes the proposed hybrid system from the
sort of petty-patent systems that exist in many countries. See Janis, supra note 112 (describing
such systems). A variant of this system is the old English caveat system, under which patents would
issue without examination unless a competitor had filed a caveat a request to be informed of any
applications filed covering certain subject matter. On the caveat system, which was used in the U.S.
during the 19th Century as well, see Malla Pollack, The Possible Unconstitutionality of Business
Method Patents: Common Sense and History 20 (2001) (working paper, on file with author).
116

30

95:#### (2001)

Rational Ignorance at the Patent Office

D. Litigation Reforms
My analysis does have significant implications for the treatment of
patent validity in litigation. Patent litigation proceeds on the assumption
that issued patents have been subjected to a thorough examination process, and that the PTO is unlikely to have made a mistake. This assumption is both explicit and implicit. The patent statute imposes a
presumption that issued patents are valid.119 The courts have read this
statute to impose a burden on the accused infringer to demonstrate invalidity of a patent by clear and convincing evidence.120 The Federal Circuit has recently strengthened that burden by backing away from an
earlier rule that applied this presumption only to arguments and prior art
actually considered during prosecution.121 The court now says the presumption exists even with respect to art that the examiner did not consider at all.122 The examiners are presumed to know their fields, and to
know the scope of the claims they allow.123 These explicit legal rules are
strengthened by jury psychology. Jurors are notoriously reluctant to second-guess patent examiners, not only because they assume examiners
know more than they do but because they believe examiners spend much
more time examining any given patent than they actually do.124
The presumption of validity has little if any basis in fact. Examiners
do not in fact spend long hours poring over a patent application or the
prior art. They spend very little time, and far less than either the lawyers
or the triers of fact in infringement cases. They regularly miss the most
relevant prior art. The Quillen-Webster study suggests, moreover, that
they ultimately issue an astonishing ninety-seven percent of the unique
applications filed.125 The law should not ignore the fact that a patent application has been examined, but it seems clear we give that examination
process far too much weight. At a bare minimum, the presumption of
validity should be pared back so that it covers only prior art references
and arguments actually considered by the examiner. There is simply no
reason to defer to the examiners purported decision on an issue the examiner did not even address. But I would go further and eliminate the
119

35 U.S.C. 282 (West Supp. 2000).


See, e.g., Al-Site Corp. v. VSI Intl., 174 F.3d 1308 (Fed. Cir. 1999) (describing the presumption of validity as strong).
121
See, e.g., Mfg. Research Corp. v. Graybar Elec. Co., 679 F.2d 1355 (11th Cir. 1982); NDM
Corp. v. Hayes Prod. Inc., 641 F.2d 1274 (9th Cir. 1981); Lee Blacksmith Inc. v. Lindsay Bros., 605
F.2d 341 (7th Cir. 1979) (adopting the considered art only rule).
122
See, e.g., Kahn v. General Motors Corp., 135 F.3d 1472, 1480 (Fed. Cir. 1998) (The presentation of evidence that was not before the examiner does not change the presumption of validity);
Applied Materials, Inc. v. Advanced Semiconductor Materials Am. Inc., 98 F.3d 1563 (Fed. Cir.
1996) (holding the same).
123
See, e.g., Al-Site, 174 F.3d at 1308 (stating that the presumption of validity carries with it a
presumption that the examiner did his duty and knew what claims he was allowing (quoting Intervet Am., Inc. v. Kee-Vet Labs., Inc., 887 F.2d 1050, 1054 (Fed. Cir. 1989))); W. Elec. v. Piezo
Tech., 860 F.2d 428 (Fed. Cir. 1988) (holding that a presumption stands regardless of the expertise
of the particular examiner in the field in question).
124
This argument is based on anecdotal evidence, but in fact there is statistical support for the
propositions that juries tend to favor patentees on validity questions and that juries are unlikely to
second-guess an examiner who has already considered and rejected a prior art reference. See Allison
& Lemley, supra note 4, at 212-13, 231-34.
125
Quillen & Webster, supra note 11, at 9 tbl.6.
120

31

NORTHWESTERN UNIVERSITY LAW REVIEW

clear and convincing evidence requirement altogether,126 making the presumption that a patent is valid rebuttable by a preponderance of the evidence.127 My argument in this article is based on the idea that it is more
efficient to decide validity after in-depth consideration in those few cases
in which it matters than to decide it upon a cursory review of all patent
applications. My argument is undermined if validity litigation does not
in fact involve a searching investigation of validity, but instead defers to
the cursory review already conducted. Based on what we know of patent
examinations, deference is not appropriate.128 If there is to be deference
to PTO decisions in litigation, it should be coupled with some form of
real third-party opposition system, and only patents that have been
through that system should be entitled to deference.
If the general statutory presumption of validity is inappropriate, the
conclusive presumption of validity temporarily afforded a pharmaceutical patent under the Hatch-Waxman Act is indefensible. As that statute
has been implemented, a prospective generic manufacturer who wants to
make a drug that may be covered by a patent files an abbreviated new
drug application (ANDA), at which point the patentee is notified. If
the patentee files an infringement lawsuit against the generic manufac-

126
Needless to say, therefore, I disagree with commentators such as Jay Kesan, who would
strengthen the presumption of validity. See Kesan, supra note 1, at 26. Kesans argument is that we
dont need validity litigation if we give patent applicants sufficient incentive to disclose the relevant
prior art in the first place. However, while such an approach might well encourage applicants to disclose prior art, the PTO could not evaluate all that prior art in a definitive way without spending
much more time in prosecution than they currently do. Thus, Kesans proposed system would likely
result in patentees effectively insulating bad patents from judicial review by overwhelming patent
examiners with prior art, knowing that the examiners wont be able to do a thorough analysis in the
time they have available.
127
This proposal is in keeping with the standard used in trademark cases. Keeping the burden
with the accused infringer takes some account of the fact that the patentee did go through some
prosecution process. It also prevents procedural anomalies from arising in cases in which an accused
infringer files for declaratory judgment that a patent is invalid.
128
A few scholars have defended deference to the PTO, but mostly in the context of appeals
from the rejection of a patent application. See, e.g., Craig Allen Nard, Deference, Defiance, and the
Useful Arts, 56 OHIO ST. L.J. 1415 (1995); Rai, supra note 63, at 202 (arguing that deference is appropriate to patent denials but not to patent grants, because the system is set up to err on the side of
granting patents). These arguments tend to focus on the proper interpretation of the Administrative
Procedures Act (APA). See generally Dickinson v. Zurko, 527 U.S. 150 (1999) (holding that the
APA requires deference to PTO factfinding).
There is some evidence that deference by courts in practice is far less than complete. About 46%
of all patents litigated to a final judgment on validity issues are held invalid. See Allison & Lemley,
supra note 4, at 205-07. This number includes decisions on appeal and at summary judgment. At
trial, only 33% of patents are held invalid. See Moore, supra note 25, at 390 Table 4. Thus, courts
are clearly making some inquiry into validity on their own. I should note, however, that the invalidity numbers have declined significantly over the past 25 years, from 65% to 46%. GLORIA K.
KOENIG, PATENT INVALIDITY: A STATISTICAL AND SUBSTANTIVE ANALYSIS 4-19 to 4-23 (rev. ed.
1980); Allison & Lemley, supra note 4, at 206 n.53. Thus, deference to the PTO by the courts seems
to be on the rise.
Finally, it is worth noting some empirical evidence that suggests that the Federal Circuits moves
to strengthen the presumption of validity had the effect of increasing both the number of patent cases
filed and the percentage that go to trial. See Jean O. Lanjouw & Josh Lerner, The Enforcement of
Intellectual Property Rights: A Survey of the Empirical Literature, 49/50 ANNALES DECONOMIE ET
DE STATISTIQUE 223 (1998). Thus, the presumption of validity may actually account for some portion of the social cost of litigation over bad patents.

32

95:#### (2001)

Rational Ignorance at the Patent Office

turer,129 the FDA cannot proceed to consider the ANDA for thirty
months, unless the Federal Circuit issues an opinion during that time
holding the patent invalid.130 In effect, the FDA acts as if the patent is
conclusively presumed valid unless the Federal Circuit instructs it otherwise. Even a district court determination of invalidity wont enable the
generic to proceed with its application as long as there is an appeal pending.131 This presumption of validity has even less basis in fact than the
litigation presumption, and its effect is to delay drug price competition
for several years even where a patent is clearly invalid. This conclusive
presumption of validity imposes significant social costs based on the assertion of bad patents in litigation, and should be abolished.
A second possible implication for litigation concerns the imposition
of costs on an accused infringer who must defend against a bad patent.
Leaving the validity determination until trial imposes significant costs on
accused infringers. Where the validity issue wasnt close, and the patent
was clearly invalid, it may make sense to shift some of those costs to the
patentee. Unfortunately, the patent law makes it very difficult for a prevailing defendant to obtain an award of attorneys fees. The statute requires the case to be exceptional.132 As a practical matter, patentees
are far more likely to be awarded attorneys fees than accused infringers,
because they can often get fees by proving willful infringement.133 By
contrast, accused infringers normally wont get attorneys fees unless
they can prove that the suit was filed in bad faith.134 It may make sense
to add some balance to the fee awards. In copyright law, for example,
the standards for prevailing parties seeking a fee award are identical
whether the party is a plaintiff or a defendant. The Supreme Court has
identified the important public policy interests not only in enforcing
copyrights but also in defending against unfounded copyright claims.135
The result of doing so has been a dramatic growth in fee awards to defendants in copyright cases.136 A similar rule might encourage more fee
shifting in patent cases, and therefore help shift some of the burden of
determining validity away from accused infringers.
Both of these suggestions would require changes in existing law.
Those changes need not be complex, however, and do not necessarily require legislative intervention. The statutory presumption of validity does
not specify that the presumption of validity can only be overcome by
129

The lawsuit would presumably proceed under 35 U.S.C. 271(e), as the generic has not yet
begun making, using or selling the product and therefore does not violate 271(a).
130
21 C.F.R. 314.107(b)(3).
131
Id.
132
35 U.S.C. 285.
133
See Great N. Corp. v. Davis Core & Pad Co., Inc., 782 F.2d 159 (Fed. Cir. 1986) (holding
that willful infringement alone can make a case exceptional and therefore justify a fee award).
134
See, e.g., Cambridge Prods., Ltd. v. Penn Nutrients, Inc., 962 F.2d 1048, 1050-51 (Fed. Cir.
1992) (In the case of awards to prevailing accused infringers . . . , exceptional cases are normally
those of bad faith litigation or those involving fraud or inequitable conduct by the patentee in procuring the patent.).
135
See Fogerty v. Fantasy, Inc., 510 U.S. 517, 527 (1994).
136
See Jeffrey Edward Barnes, Comment, Attorneys Fee Awards in Federal Copyright Litigation After Fogerty v. Fantasy: Defendants are Winning Fees More Often, but the New Standard Still
Favors Prevailing Plaintiffs, 47 UCLA L. REV. 1381 (2000).

33

NORTHWESTERN UNIVERSITY LAW REVIEW

clear and convincing evidence. The clear and convincing evidence standard is a judicial creation, and courts could read the statute to impose a
lesser burden on accused infringers. For example, 282 might be interpreted only to shift the burden of proof by preponderance of the evidence, and not to impose a higher evidentiary burden on accused
infringers.
Section 285 provides that [t]he court in exceptional cases may
award reasonable attorney fees to the prevailing party.137 Standing
alone, there is nothing in this statutory provision that suggests its application should favor one side or the other. Thus, a court could quite plausibly read it as establishing a neutral standard for fee shifting. Because of
the difficulty of proving a case exceptional in practice, however, Congress might choose to amend the statute to provide for fee shifting in
cases in which a prevailing party on either side has vindicated an important public interest.
E. Directions for Research
Finally, the analysis I have conducted suggests some directions for
future research. While many of the numbers I employ are quite definite,
or at least clear within ranges, some are not. We know very little about
how many patents are licensed for royalties. We also dont know how
intensively they are licensedthat is, how many licensees there are for
each licensed patent. As we have seen, both of these numbers are critical
to understanding the social costs of the patent system. Both seem fruitful
areas for future empirical research.
CONCLUSION
The PTO is rationally ignorant of the objective validity of the patents it examines. For the PTO to gather all the information it needs to
make real validity decisions would take an enormous investment of time
and resources. Those decisions can be made much more efficiently in
litigation, because only a tiny percentage of patents are ever litigated or
even licensed to others. Thus, we should resign ourselves to living with
a system in which bad patents do slip through the PTO undetected.
Recognizing that this is the case, however, should also prompt us to
strengthen the validity inquiry made by the trial courts. Courts should
not be ignorant of the facts, and they should not presume that a patent is
valid merely because the PTO says it is.

137

34

35 U.S.C. 285.